sensors

Article

Strain Transfer for Optimal Performanceof Sensing Sheet

Matthew Gerber Campbell Weaver Levent E Aygun ID Naveen Verma James C Sturmand Branko Glišic

Departments of Civil and Environmental Engineering and Electrical Engineering Princeton UniversityPrinceton NJ 08544 USA gerberbeamgmailcom (MG) campbellweavermaccom (CW)laygunprincetonedu (LEA) nvermaprincetonedu (NV) sturmprincetonedu (JCS) Correspondence bglisicprincetonedu

Received 28 April 2018 Accepted 5 June 2018 Published 12 June 2018

Abstract Sensing sheets based on Large Area Electronics (LAE) and Integrated Circuits (ICs) arenovel sensors designed to enable reliable early-stage detection of local unusual structural behaviorsSuch a device consists of a dense array of strain sensors patterned onto a flexible polyimide substratealong with associated electronics Previous tests performed on steel specimens equipped with sensingsheet prototypes and subjected to fatigue cracking pointed to a potential issue individual sensorsthat were on or near a crack would immediately be damaged by the crack thereby rendering themuseless in assessing the size of the crack opening or to monitor future crack growth In these testsa stiff adhesive was used to bond the sensing sheet prototype to the steel specimen Such an adhesiveprovided excellent strain transfer but it also caused premature failure of individual sensors withinthe sheet Therefore the aim of this paper is to identify an alternative adhesive that survivesminor damage yet provides strain transfer that is sufficient for reliable early-stage crack detectionA sensor sheet prototype is then calibrated for use with the selected adhesive

Keywords structural health monitoring strain transfer sensing sheet large area electronicsdamage detection flexible adhesive

1 Introduction

A critical detail in large area electronics (LAE) for damage detection is the way the sensing sheetis attached to the structure While previous research has successfully addressed this issue for sensorsbonded to less hard materials such as concrete this research explores the behavior of three newadhesive options for sensors attached to a hard material such as aluminum The most suitable adhesiveis selected by performing both load (non-destructive) and damage tests using individual full-bridgestrain sensors Factors such as set time strain transfer and adhesive strength are considered in selectingthe adhesive Using the selected adhesive an LAE strain sensor is calibrated for strain detection

11 Strain-Based Monitoring

While there are many parameters that can be monitored to detect structural damage strain isfrequently used because it is directly related to stress (eg according to Youngrsquos Modulus)Since materials fail when stress exceeds the ultimate limit state stress would be the ideal structuralparameter to monitor in order to evaluate a structurersquos health However because there are currentlyno effective means to monitor stress on a structure under real conditions strain is selected as the nextmost suitable parameter Damage to a structure (eg cracking or bowing) frequently creates strainfield anomalies that could be reliably detected if the strain sensor is installed in the perturbed area(ie in direct contact with or in close proximity to the damage) [1] The most mature and widely used

Sensors 2018 18 1907 doi103390s18061907 wwwmdpicomjournalsensors

Sensors 2018 18 1907 2 of 16

strain monitoring sensors involve resistive strain gages A strain change in the structure changes thegage wire length thus changing the wirersquos cross-sectional area and resistance A change in the wirersquosresistance causes a change in voltage drop across the sensor [2]

As an example a full-bridge resistive strain sensor that contains the Wheatstone Bridge is shownin Figure 1 While this is merely one example of a resistance-based strain sensor understanding howit works will prove to be helpful in thinking about how sensors translate a change in strain into aquantifiable measurement that allows for damage assessment See [3] for Equations (1) to (4)

Figure 1 Wheatstone Bridge The dashed line represents a crack running through the sensor [3]

The Wheatstone Bridge consists of four resistorsmdashR1 R2 R3 and R4 Resistors R1 and R3 areoriented parallel to the strain measurement axis and resistors R2 and R4 are oriented perpendicularto this axis The ratio of output voltage (Vout) to input voltage (Vin) is expressed in terms of the fourresistors and is given by Equation (1)

Vout =( R4

R3 + R4minus R1

R1 + R2

)Vin (1)

If each sensing element has the same initial resistance Equation (1) simplifies to Vout = 0 If acrack occurs at least one of the four resistive strain sensors is affected causing a change in resistance∆R From Equation (1) the relationship between change in resistance ∆R and change in voltage∆Vout is given by Equation (2)

Vout + ∆Vout = ∆Vout =( R4 + ∆R4

R3 + ∆R3 + R4 + ∆R4minus R1 + ∆R1

R1 + ∆R1 + R2 + ∆R2

)Vin (2)

The gage factor for a resistive element is defined as the ratio of relative change in resistanceto relative change in strain GF = ∆RR

ε Equation (2) can be written in terms of gage factor andstrain change

∆Vout =( 1 + GFε4

2 + GFε3 + GFε4minus 1 + GFε1

2 + GFε1 + GFε2

)Vin (3)

In the case of uni-axial stress as shown in Figure 1 resistive elements R1 and R3 experience fullstrain because they are oriented parallel to the strain axis but elements R2 and R4 also experiencestrain according to Poissonrsquos ratio ν Finally we arrive at an equation for strain change in terms ofvoltage ratio (∆Vr =

∆VoutVin

) gage factor (GF) and Poissonrsquos ratio (ν) given by Equation (4)

ε =minus2∆Vr

GF[(ν + 1)minus ∆Vr(ν minus 1)](4)

An important property about the full bridge sensor is its insensitivity to temperature changeA change in temperature of the sensor ∆T causes a change in temperature in each of the fourresistive elements ∆T = ∆TR1 = ∆TR2 = ∆TR3 = ∆TR4 because all four resistive elements are closeto each other A change in temperature ∆T causes a change in resistance ∆RT according to theequation ∆RT = Roα∆T where ∆RT = ∆RT1 = ∆RT2 = ∆RT3 = ∆RT4 Since the resistance in eachresistor after manufacturing is assumed to be the same R = R1 = R2 = R3 = R4 Equation (2) yields

Sensors 2018 18 1907 3 of 16

∆Vout = 0 so a change in temperature does not cause a change in voltage Therefore temperaturecompensation is not necessary in this type of sensor

12 2D Strain Monitoring

Besides resistive strain sensors there are several other types of strain sensors based either onthe physical principle behind measurement (eg vibrating wires fiber optics etc) or on the spatialextent of sensors Based on the spatial extent of sensors the sensors could be discrete short-gagelong-gage distributed one-dimensional (1D) or two-dimensional (2D) [1] Regardless of the type orspatial extent strain sensors that are in direct contact with the damage-induced strain-field anomalycan detect this anomaly reliably as the anomaly will cause an unusually high change in output signalof the sensor [4] Nevertheless the spatial extent of the sensor plays an important role in increasing ordecreasing the likelihood of detection of strain-field anomalies (ie of damage) Figure 2 demonstratesthe damage detection capabilities for short-gage long-gage distributed and two-dimensional sensorsAs expected the distributed sensor performs the best among the three one-dimensional options but isunable to detect damage of type E since the damage lies out of reach for the sensor The 2D sensoron the other hand promises to reliably detect all four types of damage illustrated in this example as itis in direct contact with each of them

Figure 2 Short-gage long-gage distributed and 2D sensor damage detection capabilities [1]

13 2D Strain Sensing Research and Testing

Discrete long-gage and short-gage sensors as well as 1D distributed sensors are commerciallyavailable which is not the case for 2D distributed sensors Great promise for reliable damage detectionmotivates the need for research in that area Hence several researchers have explored 2D sensors invarious forms of sensing skins [5ndash9] expandable sensor networks [10] nano-paints and nano-basedmaterials and adhesives [11ndash14] photonic crystals [15] etc

Several researchers from Princeton University jointly created a 2D sensor called sensing sheetbased on Large Area Electronics (LAE) and Integrated Circuits (ICs) [1ndash316ndash20] LAE enables theintegration of dense arrays of diverse sensors and accessory electronics onto thin deformable plasticsubstrates which can cover large areas of structures and enable direct damage [21ndash25] Sensing sheetshave three main components [16] (1) a two-dimensional dense array of unit strain sensors patternedon a polyimide (Kapton) substrate and combined with functional LAE (2) embedded ICs interfacedvia non-contact links for sensor readout data analysis power management and communicationand (3) an integrated flexible photovoltaic sheet and power converters (rechargeable batteries) onthe LAE sheet to power the full system and protect it from elements (wind rain snow ultravioletradiation etc)

Sensors 2018 18 1907 4 of 16

It is important to note that while the general sensing principle of sensing sheet is based onstrain measurement its purpose is not to accurately measure strain magnitudes but rather to detectlocalize and quantify unusual behaviors that manifest in the form of strain-field anomalies (egcracking of concrete fatigue cracking and bowing of steel) Being in direct contact with sensing sheetindividual sensors the strain-field anomalies create unusually high strain changes in these sensorscompared with individual sensors that are not in direct contact with the anomaly This unusually highstrain change is used to directly detect and localize the anomaly The extent of the anomaly is theninferred from geographical locations of affected individual sensors This principle is shown in Figure 3

The applications of sensing sheet could be on a variety of civil and mechanical structures andinfrastructure including but not limited to bridges buildings tunnels pipelines wind turbinesaircrafts etc Given that accurate measurement of strain is not required the imperfections andphenomena that affect measurements in the range of usual strain changes (eg potential rheologicalstrain in glue lower strain transfer due to gluing over the paint of original structure etc) wouldnot significantly affect the performance of the sensing sheet for this function More detail is given inliterature [1ndash316ndash20]

Figure 3 Schematic representation of the sensing sheet based on LAE and ICs [3]

Yao and Glišic [16] performed preliminary tests to evaluate whether an early-stage LAE prototypehas the potential to detect damage such as a crack [1] The authors tested a full-bridge resistance-basedstrain sensor as a preliminary step towards LAE

Series of tests were performed at the Carleton Laboratory at Columbia University on an LAE-basedstrain sensing sheet prototype in order to evaluate the sensorrsquos ability to detect cracks in steel platespecimens [16] In these tests the sensors were bonded to the steel samples with a notch and wereexposed to fatigue cycling so that the crack occurrence and propagation could be predicted While thetests were successful from the point of view of damage detection they revealed a challenge individualsensors would fail immediately upon damage occurrence thereby limiting the ability of sensing sheetsto follow and quantify damage progression That challenge is addressed in this paper

Sensors 2018 18 1907 5 of 16

2 Selecting a Suitable Adhesive

21 Objective

Because manufacturing sensing sheets at the prototype stage is expensive the following tests areperformed using existing identical strain-gages

A critical component of a sensorrsquos functionality is the way the sensor is fixed to the structurersquossurface Because the sensing sheets involved in this study are formed on a polyimide substratean adhesive is the best fixing option At minimum a suitable adhesive should maintain bond betweenthe sensor and structure under various loads damage types and environmental conditions An equallyimportant consideration involves the adhesiversquos elastic behavior particularly in materials that do notmechanically degrade when damaged Due to the heterogeneous nature and particle flaws concretedegrades when subjected to tensile stresses When subjecting a strain sensor bonded to concreteto moderate strains the bond between the sensor and concrete is stronger than the bond betweenaggregate particles in the concrete causing the concrete to break apart near the crack In this case thestrain in the sensor is defined over the strain length shown in Figure 4 Conversely when subjectinga strain sensor bonded to non-degrading materials such as aluminum and steel the bond betweenparticles in the material is much stronger than the bond between the sensing sheet and structure sothe strain is defined over the sample separation (See Figure 5) leading to much higher strain valuesMechanical strain in the sensor is calculated as the change in length divided by original length Fora perfectly non-degrading material that has no initial separation any separation will theoreticallycause infinite strain because original separation length is zero Hence in non-degrading materials thesensor will fail soon after crack occurrence which in turn will disable monitoring of crack size and itsevolution Therefore the purpose of this study is to identify and both qualitatively and quantitativelyevaluate an adhesive to be used for two-dimensional sensing on non-degrading materials so that thesensor (sensing sheet) survives crack occurrence and continues monitoring the crack size

The soft adhesives considered in this research are expected to have viscoelastic behaviorand consequently rheological effects in adhesives may influence values of strain measurements overtime However as the main aim of this research was to identify an adhesive that enables the sensingsheet to survive the damage to the structure only short-term effects were analyzed and long-termeffects will be a subject of future work This approach was considered suitable as the maximum stressin sensing sheet happens instantaneously with occurrence of damage (rheological effects are actuallybeneficial as they bring relaxation) In addition rheological effects are small compared to strain changesdue to damage and thus they do not significantly affect its damage detection capability (they affectstrain measurement but accurate strain measurement is not the purpose of the sensing sheet)

Figure 4 Strain length vs sample separation on degrading materials Degradation is indicated byred lines

Sensors 2018 18 1907 6 of 16

Figure 5 Strain length vs sample separation on non-degrading materials

22 Qualitative Tests

Qualitative observations were performed for three high-strength flexible adhesive alternatives3M DP100 DP105 and DP125 Initially these three adhesives were tested on paper samples in orderto gain an understanding of viscosity flexibility and set time It was found that the three adhesiveslisted were very flexible and could be bent and twisted (along with the substrate paper sheet) byhand without much effort A sample of a flexible adhesive applied to paper and bent can be seenin Figure 6 below

Figure 6 Sample of a flexible adhesive applied to paper

The 3M DP100 had a set time of about five minutes the DP105 had a set time of about 20 minutesand the DP125 had a set time which exceeded half an hour An ideal set time is neither too short(to give adequate time to place the sensor sheet) nor too long (to avoid sensor displacement caused byaccidental contact shortly after placement) A set time exceeding a half hour is undesirable because itincreases the risk of accidentally moving or removing the sensor during the installation phase Becausethe DP125 adhesive set time exceeded half an hour it was eliminated and the DP100 and DP105adhesives were further tested

Sensors 2018 18 1907 7 of 16

23 Quantitative Tests Loading and Unloading

Strain tests were performed in order to understand how much strain is transferred to the sensorfrom a monitored material given that the adhesives are soft and some loss of strain transfer is expectedThe test involved bonding three sensors to the top of a thin aluminum simply supported beam andloading and unloading the beam with known point forces Due to limitations of available testingset-up bottles filled with sand were used to load the beam The beam was simply supported at bothends as shown in Figure 7 It had span of 1710 mm thickness of 913 mm and width of 254 mmThe beam was made of aluminum and had a Youngrsquos modulus of 68 GPa Strain gages were placed at855 mm from the left support (in the middle of the beam) The sand bottles B1 B2 and B3 had masses5942 g (58 kN) 5981 g (59 kN) and 5977 g (59 kN) respectively For each load step the strain valuewas recorded with an Omega DP41-S reading unit (Omega Engineering Stamford CT USA) The totalloading and unloading time was less than 2 minutes The three tested adhesives were a stiff Araldite2012 adhesive (used by Yao and Glišic in previous studies [16]) 3M DP100 and 3M DP105 The stiffadhesive was used for reference purposes as it is assumed to have high quantity of strain transferdue to its high stiffness The aluminum beam and three bonded sensors used in the test are shown inFigure 7 below

Figure 7 Strain transfer test set-up (a) view of beam without sensors (b) support detail (c) view ofsensors installed using Araldite 2012 3M DP 100 and 3M DP 105 (d) sensor details

The results of the strain transfer tests are shown in Figure 8 An analytic strain model wasdeveloped for the aluminum beam The model was based on the linear theory of beams and ispresented in Equation (5)

εs =M times ys

EI(5)

where M = FL24mdashbending moment at location of the sensors (F is force applied using bottles placedin the middle of the beam L = 1710 mm is length of the beam)

E = 68 GPamdashYoungrsquos modulus of aluminumI = bh312mdashmoment of inertia of the beamrsquos cross-section (b = 254 mm h = 913 mm)

and ys = h2 + 02 mmmdashdistance of sensors from the horizontal principal axis of the cross-section(where 02 mm accounts for estimated thickness of the glue plus half the thickness of the strain gage)

Sensors 2018 18 1907 8 of 16

Figure 8 shows the instantaneous bending strain for seven load steps The first and last load stepscorrespond with no beam loading the second and sixth load steps correspond with the beam loadedwith sand bottle B1 the third and fifth load steps correspond with the beam loaded with sand bottlesB1 and B2 and the fourth load step corresponds with the beam loaded with all three sand bottlesAs expected the stiff Araldite adhesive consistently transfers the greatest fraction of strain from thestructure to the sensor and none of the adhesives transferred all strain to the sensor The strain valuesmeasured by sensors glued with Araldite were about 25 times greater than those measured by theother two glues Figure 8 shows that while DP100 transfers slightly more strain than DP105 in generalthere is no significant strain transfer difference between the 3M DP100 and 3M DP105 adhesivesBoth adhesives ldquoloserdquo more than 50 percent of strain however this is acceptable as the main purposeof the sensing sheet is damage detection and not strain monitoring Moreover linear response ofboth adhesives indicates that lack in strain transfer can be accounted for through calibration and gagefactor adjustments This is effectively done in Section 3 of this paper where strain transfer was testedon sensing sheet prototypes

Figure 8 Strain transfer results for various adhesives

24 Damage Tests Crack Opening vs Strain Test

The aim of the damage test was to assess the behavior of adhesives at crack locations andto see how it affects the survival of the sensors For these tests a sensor was bonded to ldquoclosedrdquoaluminum plates and the adhesive was given time to cure Once the adhesive reached suitablestrength an artificial crack was generated by pulling the aluminum plates apart thereby creating a gapbetween them The test setup is shown in Figure 9 Each simulated crack opening increment involvedpulling the aluminum plates apart by 0001 inches taking the strain reading and then waiting fiveminutes to record the ldquorelaxedrdquo strain reading Relaxation is expected due to creep in sensor polyimidesubstrate as well as due to creep in the adhesive exposed to high stress levels This process wasrepeated until failure (ie until the sensor could no longer read the strain due to rupture or the strainvalues began to decrease due to debonding or the adhesive breaking) The tests were performed withVishay 3800 strain reading units and the gage factor was calibrated to the manufacturerrsquos specificationsAn example of post-test damage is shown below in Figure 9

Sensors 2018 18 1907 9 of 16

Figure 9 Crack test setup

Based on the manufacturerrsquos specifications both adhesives achieve 80 of their full strengths ifcured for 20 minutes at 23 C and they achieve full strength if cured for 48 hours at 23 C Given thatthe temperature of the lab in which the tests were made was set to 17 C and could not be brought upto 23 C it was decided to perform one test before the full strength was achieved and one after the fullstrength was achieved For practical reasons the former was performed 24 hours after gluing and thelatter 72 hours after gluing (one day longer than specified time of 48 hours in order to account forlower curing temperature)

The following plots show the relationship between crack opening size and strain reading forthe 3M DP100 adhesive Both of these tests were performed with strain gages with a gage factor of201 In the first test (Figure 10) the adhesive was allowed to set for 24 hours and in the second testfor 72 hours (Figure 11) Time-dependent tests were performed in order to evaluate the adhesivesrsquoperformance with time and to establish a minimum cure time for future lab tests

Figure 10 DP100 crack opening vs strain after 24 hours

Sensors 2018 18 1907 10 of 16

Figure 11 DP100 crack opening vs strain after 72 hours

The time-dependent tests for the 3M DP100 adhesive showed very interesting results The testsdemonstrated that the bond strength does show time-dependence The 24-hour DP100 test showedreadings up to a crack opening of about 015 mm whereas the 72-hour test showed readings up toa 026 mm crack The result is important because it verifies that the adhesive performs better whengiven a few days to cure

Similar tests were performed for the 3M DP105 adhesive The results are shown inFigures 12 and 13 below

Figure 12 DP105 crack opening vs strain after 24 hours

Sensors 2018 18 1907 11 of 16

Figure 13 DP105 crack opening vs strain after 72 hours

The DP105 adhesive continued to read strain for crack openings up to 045 mm compared to about026 mm for the DP100 adhesive Taking into account that all other properties of the two adhesivesare similar the DP105 adhesive is considered the more suitable adhesive for the installation of sensingsheets that are supposed to detect and survive structural damage

3 Application to a Sensing Sheet Prototype

31 Objective

Once the appropriate adhesive for the sensing sheet was selected its application was testedon a sensing sheet prototype Given that manufacturing of a sensing sheet at the prototype stageis expensive it was decided to perform only non-damage strain transfer tests so that the sensing sheetwould not be damaged and could be used for future research related to hardware

The first step in implementing the sensing sheet is to determine appropriate gage factors of theindividual sensors and observe how these gage factors change depending on the adhesive used tobond the sheet The same testing setup as described in Section 23 was used The sensing sheet wasinstalled in the third quarter of the beam ie 1282 mm from the left support The load was applied1069 mm from the left support

The prototype of the sensing sheet is composed of eight individual sensors each composed offour resistors forming a full bridge Each individual sensor was independently tested by bonding thesensing sheet to an aluminum beam then loading and unloading the beam with sand bottles similarto the strain-transfer test described in the previous section In total two sensing sheets were testedthe first glued with the stiff Araldite adhesive for reference purposes and the second glued with thesoft DP105 glue Each loading step lasted 15 s and values were sampled at 1000 Hz 10 seconds intoeach loading step The beam was loaded for seven loading steps no load B1 B1 and B2 B1 and B2and B3 B1 and B2 B1 and no load giving a total testing time of 105 seconds At each loading step thechange in output voltage was recorded Using the change in output voltage the input voltage and ananalytical strain model the gage factor was determined for each sensor under each loading conditionFor each sensor identical tests were performed three times yielding three gage factors Preliminarytests included air and beam surface temperature measurements but no significant temperature changeswere found during the test duration Figures 14 and 15 show the test setup

Sensors 2018 18 1907 12 of 16

Figure 14 Sensing sheet and individual sensor numbering convention

Figure 15 Gage factor test setup The sensing sheet is bonded to the top surface of the aluminum beamand is connected to the rigid printed circuit board The temperature reading device is shown in thebottom left corner

32 Results

For each loading step for each individual sensor apparent gage factors were determined underidentical loading conditions and they are shown in Figure 16 Note that these gage factors are apparentand not true as they combine true gage factor with strain transfer quality The figure shows thatconsistent results are obtained for all individual sensors within the same sheet In addition the gagefactors for sensors attached using the DP105 adhesive are about 25 times less than those for the sensorsattached using Araldite (see Table 1) This result is consistent with strain transfer test results describedearlier in Section 2

As expected smaller loads yielded greater gage factor errors than large loads due to lower signalto noise ratios This created dispersion in the results Figure 17 plots the gage factor uncertaintymeasured as standard deviation for each load step Indices 1 and 5 represent the beam loaded withbottle B1 2 and 4 represent loading with bottles B1 and B2 and 3 represents loading with all three

Sensors 2018 18 1907 13 of 16

bottles Loading step 3 has the smallest gage factor uncertainty with a standard deviation of about006 for 3M DP105 and 013 for Araldite Loading step 5 has the greatest gage factor uncertaintywith a standard deviation of about 033 for 3M DP105 and 054 for Araldite As Figure 17 showsthe Araldite and 3M DP105 have similar gage factor uncertainty trends except the Aralditersquos gagefactor uncertainties are consistently greater than those for 3M DP105 The reason for this could be theimperfections and connection issues in the flexible to rigid interface where a zero-insertion-force (ZIF)connector is used Unwanted impedance in this connection can vary by insertion and introduces asmall amount of noise into the differential voltage which is used to calculate strain This also showsthe need for an integrated system implementation on a flexible substrate and the need to reduce thenumber of interfaces to rigid boards as mentioned in previous works [18ndash20]

Figure 16 Apparent gage factors determined on individual sensors of sensing sheet prototypes gluedwith Araldite and 3M DP105 adhesives

Figure 17 Apparent gage factor uncertainties for various load steps

Sensors 2018 18 1907 14 of 16

Table 1 Apparent gage factor summary for Araldite and 3M DP105 adhesives (microlowast = meanσlowast = standard deviation)

Araldite GF 3M DP105 GF

microlowast 135 054σlowast 0240 0128

33 Discussion

The apparent gage factor for Araldite is greater than the apparent gage factor for 3M DP105adhesive because the Araldite is less flexible and transfers more strain from the structure to the sensorStrain is proportional to change in output voltage and inversely proportional to gage factor so if ananalytical strain model is used to determine gage factor a stiff adhesive will result in a larger changein output voltage and the gage factor will be higher Obtained results are consistent with the results ofthe strain transfer test presented in Section 2

For the sensing sheet bonded with Araldite sensors 2 and 7 showed high voltage output noise inthe change in output voltage values and therefore did not yield reasonable gage factor values For thesensing sheet bonded with 3M DP105 sensor 2 had high noise and did not yield reasonable gagefactor values Because change in output voltage values for sensor 2 were unreasonable in both theAraldite and 3M DP105 tests it is likely that there is a design or manufacturing flaw in the sensingsheet itself or a soldering problem on the rigid printed circuit board (see Figure 15) In the case ofsensor 7 in the Araldite-bonded sensor sheet the high noise can likely be attributed to a manufacturingflaw on the sensor sheet itself

Figure 16 shows that there is a range in apparent gage factor values (dispersion) from sensorto sensor One reason the gage factors might vary from sensor to sensor has to do with a possiblevoltage drop from the sensor to the rigid printed circuit board and reading unit Because the wiresin the sensing sheet have high resistance sensors that are located farthest from the reading unit(sensors 0 and 4 for example) may show a lower voltage output change than sensors that are closerto the reading unit (sensors 3 and 8) Furthermore a lower voltage output change will result in alower gage factor Figure 16 shows generally lower gage factors for sensors 0 and 4 validating thishypothesis Sensors 1 and 5 show higher gage factors and exhibit similar behavior Sensor 7 doesnot follow the trend described which may be attributed to a manufacturing flaw Another reasonfor varying gage factors between sensors is that the adhesive layerrsquos thickness might vary across thesensor sheet A thinner adhesive layer will permit more strain to be transferred to the sensor and thegage factor will be higher Conversely a thicker adhesive layer will transfer less strain to the sensorand will result in a lower computed gage factor

Finally Figure 17 and Table 1 show that the Araldite-bonded sensing sheet yields greater gagefactor uncertainties than the 3M DP105-bonded sensing sheet The Araldite adhesive is much moreviscous than the 3M DP105 adhesive and did not spread as easily when bonding the sensor sheet to thealuminum beam resulting in minor air pockets under the sensor sheet With the 3M DP105 adhesiveit was much easier to spread an even layer on the aluminum beam and no air pockets were visiblelikely increasing apparent gage factor consistency both between and within sensor sheets

4 Conclusions

The aim of this paper was to determine a suitable adhesive for installation of a novel sensing sheetonto steel structural members subjected to cracking Previous research has shown that for sensingsheets installed using a stiff adhesive (eg Araldite) the cracking in steel would induce immediatefailure of individual sensors of the sheet To address this challenge three soft 3M glues were consideredDP100 DP105 and DP125

The 3M DP125 adhesive was eliminated because the set time was too long to be useful inpractical applications Strain transfer tests were performed using strain gages to assess the quality

Sensors 2018 18 1907 15 of 16

of strain transfer of the two other glues It was shown that while the Araldite adhesive transfersa large percentage of strain to the sensors bonding with both the 3M DP100 and DP105 providessufficient sensitivity to detect small strain changes even in the non-damage range No conclusiveresults were found to justify the preference between the DP100 or DP105 from the non-damage studyso crack opening damage tests were performed 48-hour and 72-hour cure tests showed that the DP105adhesive was able to detect strain changes for larger crack opening sizes compared to the DP100 so itwas deemed more suitable for damage detection (cracking) in mechanically non-degrading materials(eg metals)

A sensing sheet prototype manufactured at Princeton University was then tested using the 3MDP105 adhesive For each of the working individual sensors apparent gage factors were determinedby loading and unloading a thin aluminum beam with sand bottles using theoretical strain valuesunder Euler-Bernoulli beam theory Uncertainties in the apparent gage factor values were found forboth Araldite and 3M DP105-bonded sensor sheets As expected the apparent gage factors for theAraldite adhesive were higher than for the 3M DP105 adhesive We know that an approximate gagefactor for bonding a sensor sheet with 3M DP105 adhesive would allow one to use the sensor sheet toidentify real strain values on structures

The consistency between the strain transfer tests performed on regular strain gages and the testsperformed on sensing sheets confirm that DP105 can be used for the installation of sensing sheets onsteel structural members for crack detection and characterization purposes

Author Contributions BG NV and JES created the concept of sensing sheet MG designed and carried outthe tests and data analysis he is the lead author of the paper CW and LEA developed and manufacturedsensing sheet prototype and associated hardware

Acknowledgments This research was in part supported by the Princeton Institute for the Science andTechnology of Materials (PRISM) and in part by the USDOT-RITA UTC Program grant no DTRT12-G- UTC16enabled through the Center for Advanced Infrastructure and Transportation (CAIT) at Rutgers UniversityThe authors would like to thank Y Hu L Huang N Lin W Rieutort-Louis J Sanz-Robinson T Liu S Wagnerand J Vocaturo from Princeton University for their precious help Kaitlyn Kliever helped improve the contents ofthis paper

Conflicts of Interest The authors declare no conflict of interest

References

1 Yao Y Glišic B Reliable damage detection and localization using direct strain sensing In Proceedings ofthe Sixth International IAMBAS Conference on Bridge Maintenance Safety and Management Stresa Italy8ndash12 July 2012 pp 714ndash721

2 Yao Y Tung S-TE Glišic B Crack detection and characterization techniquesmdashAn overview Struct ControlHealth Monit 2014 21 1387ndash1413 [CrossRef]

3 Tung S-T Yao Y Glišic B Sensing Sheet The Sensitivity of Thin-Film Full- Bridge Strain Sensors for CrackDetection and Characterization Meas Sci Technol 2014 25 14 [CrossRef]

4 Glišic B Chen J Hubbell D Streicker Bridge A comparison between Bragg-gratings long-gaugestrain and temperature sensors and Brillouin scattering-based distributed strain and temperature sensorsIn Proceedings of the SPIE Smart StructuresNDE and Materials + Nondestructive Evaluation and HealthMonitoring San Diego CA USA 6ndash10 March 2011

5 Loh KJ Hou TC Lynch JP Kotov NA Carbon Nanotube Sensing Skins for Spatial Strain and ImpactDamage Identification J Nondestruct Eval 2009 28 9ndash25 [CrossRef]

6 Pyo S Loh KJ Hou TC Jarva E Lynch JP A wireless impedance analyzer for automated tomographicmapping of a nanoengineered sensing skin Smart Struct Syst 2011 8 139ndash155 [CrossRef]

7 Pour-Ghaz M Weiss J Application of Frequency Selective Circuits for Crack Detection in ConcreteElements J ASTM Int 2011 8 1ndash11

8 Schumacher T Thostenson ET Development of structural carbon nanotube-based sensing composites forconcrete structures J Intell Mater Syst Struct 2014 25 1331ndash1339 [CrossRef]

Sensors 2018 18 1907 3 of 16

∆Vout = 0 so a change in temperature does not cause a change in voltage Therefore temperaturecompensation is not necessary in this type of sensor

12 2D Strain Monitoring

Besides resistive strain sensors there are several other types of strain sensors based either onthe physical principle behind measurement (eg vibrating wires fiber optics etc) or on the spatialextent of sensors Based on the spatial extent of sensors the sensors could be discrete short-gagelong-gage distributed one-dimensional (1D) or two-dimensional (2D) [1] Regardless of the type orspatial extent strain sensors that are in direct contact with the damage-induced strain-field anomalycan detect this anomaly reliably as the anomaly will cause an unusually high change in output signalof the sensor [4] Nevertheless the spatial extent of the sensor plays an important role in increasing ordecreasing the likelihood of detection of strain-field anomalies (ie of damage) Figure 2 demonstratesthe damage detection capabilities for short-gage long-gage distributed and two-dimensional sensorsAs expected the distributed sensor performs the best among the three one-dimensional options but isunable to detect damage of type E since the damage lies out of reach for the sensor The 2D sensoron the other hand promises to reliably detect all four types of damage illustrated in this example as itis in direct contact with each of them

Figure 2 Short-gage long-gage distributed and 2D sensor damage detection capabilities [1]

13 2D Strain Sensing Research and Testing

Discrete long-gage and short-gage sensors as well as 1D distributed sensors are commerciallyavailable which is not the case for 2D distributed sensors Great promise for reliable damage detectionmotivates the need for research in that area Hence several researchers have explored 2D sensors invarious forms of sensing skins [5ndash9] expandable sensor networks [10] nano-paints and nano-basedmaterials and adhesives [11ndash14] photonic crystals [15] etc

Several researchers from Princeton University jointly created a 2D sensor called sensing sheetbased on Large Area Electronics (LAE) and Integrated Circuits (ICs) [1ndash316ndash20] LAE enables theintegration of dense arrays of diverse sensors and accessory electronics onto thin deformable plasticsubstrates which can cover large areas of structures and enable direct damage [21ndash25] Sensing sheetshave three main components [16] (1) a two-dimensional dense array of unit strain sensors patternedon a polyimide (Kapton) substrate and combined with functional LAE (2) embedded ICs interfacedvia non-contact links for sensor readout data analysis power management and communicationand (3) an integrated flexible photovoltaic sheet and power converters (rechargeable batteries) onthe LAE sheet to power the full system and protect it from elements (wind rain snow ultravioletradiation etc)

Sensors 2018 18 1907 4 of 16

It is important to note that while the general sensing principle of sensing sheet is based onstrain measurement its purpose is not to accurately measure strain magnitudes but rather to detectlocalize and quantify unusual behaviors that manifest in the form of strain-field anomalies (egcracking of concrete fatigue cracking and bowing of steel) Being in direct contact with sensing sheetindividual sensors the strain-field anomalies create unusually high strain changes in these sensorscompared with individual sensors that are not in direct contact with the anomaly This unusually highstrain change is used to directly detect and localize the anomaly The extent of the anomaly is theninferred from geographical locations of affected individual sensors This principle is shown in Figure 3

The applications of sensing sheet could be on a variety of civil and mechanical structures andinfrastructure including but not limited to bridges buildings tunnels pipelines wind turbinesaircrafts etc Given that accurate measurement of strain is not required the imperfections andphenomena that affect measurements in the range of usual strain changes (eg potential rheologicalstrain in glue lower strain transfer due to gluing over the paint of original structure etc) wouldnot significantly affect the performance of the sensing sheet for this function More detail is given inliterature [1ndash316ndash20]

Figure 3 Schematic representation of the sensing sheet based on LAE and ICs [3]

Yao and Glišic [16] performed preliminary tests to evaluate whether an early-stage LAE prototypehas the potential to detect damage such as a crack [1] The authors tested a full-bridge resistance-basedstrain sensor as a preliminary step towards LAE

Series of tests were performed at the Carleton Laboratory at Columbia University on an LAE-basedstrain sensing sheet prototype in order to evaluate the sensorrsquos ability to detect cracks in steel platespecimens [16] In these tests the sensors were bonded to the steel samples with a notch and wereexposed to fatigue cycling so that the crack occurrence and propagation could be predicted While thetests were successful from the point of view of damage detection they revealed a challenge individualsensors would fail immediately upon damage occurrence thereby limiting the ability of sensing sheetsto follow and quantify damage progression That challenge is addressed in this paper

Sensors 2018 18 1907 5 of 16

2 Selecting a Suitable Adhesive

21 Objective

Because manufacturing sensing sheets at the prototype stage is expensive the following tests areperformed using existing identical strain-gages

A critical component of a sensorrsquos functionality is the way the sensor is fixed to the structurersquossurface Because the sensing sheets involved in this study are formed on a polyimide substratean adhesive is the best fixing option At minimum a suitable adhesive should maintain bond betweenthe sensor and structure under various loads damage types and environmental conditions An equallyimportant consideration involves the adhesiversquos elastic behavior particularly in materials that do notmechanically degrade when damaged Due to the heterogeneous nature and particle flaws concretedegrades when subjected to tensile stresses When subjecting a strain sensor bonded to concreteto moderate strains the bond between the sensor and concrete is stronger than the bond betweenaggregate particles in the concrete causing the concrete to break apart near the crack In this case thestrain in the sensor is defined over the strain length shown in Figure 4 Conversely when subjectinga strain sensor bonded to non-degrading materials such as aluminum and steel the bond betweenparticles in the material is much stronger than the bond between the sensing sheet and structure sothe strain is defined over the sample separation (See Figure 5) leading to much higher strain valuesMechanical strain in the sensor is calculated as the change in length divided by original length Fora perfectly non-degrading material that has no initial separation any separation will theoreticallycause infinite strain because original separation length is zero Hence in non-degrading materials thesensor will fail soon after crack occurrence which in turn will disable monitoring of crack size and itsevolution Therefore the purpose of this study is to identify and both qualitatively and quantitativelyevaluate an adhesive to be used for two-dimensional sensing on non-degrading materials so that thesensor (sensing sheet) survives crack occurrence and continues monitoring the crack size

The soft adhesives considered in this research are expected to have viscoelastic behaviorand consequently rheological effects in adhesives may influence values of strain measurements overtime However as the main aim of this research was to identify an adhesive that enables the sensingsheet to survive the damage to the structure only short-term effects were analyzed and long-termeffects will be a subject of future work This approach was considered suitable as the maximum stressin sensing sheet happens instantaneously with occurrence of damage (rheological effects are actuallybeneficial as they bring relaxation) In addition rheological effects are small compared to strain changesdue to damage and thus they do not significantly affect its damage detection capability (they affectstrain measurement but accurate strain measurement is not the purpose of the sensing sheet)

Figure 4 Strain length vs sample separation on degrading materials Degradation is indicated byred lines

Sensors 2018 18 1907 6 of 16

Figure 5 Strain length vs sample separation on non-degrading materials

22 Qualitative Tests

Qualitative observations were performed for three high-strength flexible adhesive alternatives3M DP100 DP105 and DP125 Initially these three adhesives were tested on paper samples in orderto gain an understanding of viscosity flexibility and set time It was found that the three adhesiveslisted were very flexible and could be bent and twisted (along with the substrate paper sheet) byhand without much effort A sample of a flexible adhesive applied to paper and bent can be seenin Figure 6 below

Figure 6 Sample of a flexible adhesive applied to paper

The 3M DP100 had a set time of about five minutes the DP105 had a set time of about 20 minutesand the DP125 had a set time which exceeded half an hour An ideal set time is neither too short(to give adequate time to place the sensor sheet) nor too long (to avoid sensor displacement caused byaccidental contact shortly after placement) A set time exceeding a half hour is undesirable because itincreases the risk of accidentally moving or removing the sensor during the installation phase Becausethe DP125 adhesive set time exceeded half an hour it was eliminated and the DP100 and DP105adhesives were further tested

Sensors 2018 18 1907 7 of 16

23 Quantitative Tests Loading and Unloading

Strain tests were performed in order to understand how much strain is transferred to the sensorfrom a monitored material given that the adhesives are soft and some loss of strain transfer is expectedThe test involved bonding three sensors to the top of a thin aluminum simply supported beam andloading and unloading the beam with known point forces Due to limitations of available testingset-up bottles filled with sand were used to load the beam The beam was simply supported at bothends as shown in Figure 7 It had span of 1710 mm thickness of 913 mm and width of 254 mmThe beam was made of aluminum and had a Youngrsquos modulus of 68 GPa Strain gages were placed at855 mm from the left support (in the middle of the beam) The sand bottles B1 B2 and B3 had masses5942 g (58 kN) 5981 g (59 kN) and 5977 g (59 kN) respectively For each load step the strain valuewas recorded with an Omega DP41-S reading unit (Omega Engineering Stamford CT USA) The totalloading and unloading time was less than 2 minutes The three tested adhesives were a stiff Araldite2012 adhesive (used by Yao and Glišic in previous studies [16]) 3M DP100 and 3M DP105 The stiffadhesive was used for reference purposes as it is assumed to have high quantity of strain transferdue to its high stiffness The aluminum beam and three bonded sensors used in the test are shown inFigure 7 below

Figure 7 Strain transfer test set-up (a) view of beam without sensors (b) support detail (c) view ofsensors installed using Araldite 2012 3M DP 100 and 3M DP 105 (d) sensor details

The results of the strain transfer tests are shown in Figure 8 An analytic strain model wasdeveloped for the aluminum beam The model was based on the linear theory of beams and ispresented in Equation (5)

εs =M times ys

EI(5)

where M = FL24mdashbending moment at location of the sensors (F is force applied using bottles placedin the middle of the beam L = 1710 mm is length of the beam)

E = 68 GPamdashYoungrsquos modulus of aluminumI = bh312mdashmoment of inertia of the beamrsquos cross-section (b = 254 mm h = 913 mm)

and ys = h2 + 02 mmmdashdistance of sensors from the horizontal principal axis of the cross-section(where 02 mm accounts for estimated thickness of the glue plus half the thickness of the strain gage)

Sensors 2018 18 1907 8 of 16

Figure 8 shows the instantaneous bending strain for seven load steps The first and last load stepscorrespond with no beam loading the second and sixth load steps correspond with the beam loadedwith sand bottle B1 the third and fifth load steps correspond with the beam loaded with sand bottlesB1 and B2 and the fourth load step corresponds with the beam loaded with all three sand bottlesAs expected the stiff Araldite adhesive consistently transfers the greatest fraction of strain from thestructure to the sensor and none of the adhesives transferred all strain to the sensor The strain valuesmeasured by sensors glued with Araldite were about 25 times greater than those measured by theother two glues Figure 8 shows that while DP100 transfers slightly more strain than DP105 in generalthere is no significant strain transfer difference between the 3M DP100 and 3M DP105 adhesivesBoth adhesives ldquoloserdquo more than 50 percent of strain however this is acceptable as the main purposeof the sensing sheet is damage detection and not strain monitoring Moreover linear response ofboth adhesives indicates that lack in strain transfer can be accounted for through calibration and gagefactor adjustments This is effectively done in Section 3 of this paper where strain transfer was testedon sensing sheet prototypes

Figure 8 Strain transfer results for various adhesives

24 Damage Tests Crack Opening vs Strain Test

The aim of the damage test was to assess the behavior of adhesives at crack locations andto see how it affects the survival of the sensors For these tests a sensor was bonded to ldquoclosedrdquoaluminum plates and the adhesive was given time to cure Once the adhesive reached suitablestrength an artificial crack was generated by pulling the aluminum plates apart thereby creating a gapbetween them The test setup is shown in Figure 9 Each simulated crack opening increment involvedpulling the aluminum plates apart by 0001 inches taking the strain reading and then waiting fiveminutes to record the ldquorelaxedrdquo strain reading Relaxation is expected due to creep in sensor polyimidesubstrate as well as due to creep in the adhesive exposed to high stress levels This process wasrepeated until failure (ie until the sensor could no longer read the strain due to rupture or the strainvalues began to decrease due to debonding or the adhesive breaking) The tests were performed withVishay 3800 strain reading units and the gage factor was calibrated to the manufacturerrsquos specificationsAn example of post-test damage is shown below in Figure 9

Sensors 2018 18 1907 9 of 16

Figure 9 Crack test setup

Based on the manufacturerrsquos specifications both adhesives achieve 80 of their full strengths ifcured for 20 minutes at 23 C and they achieve full strength if cured for 48 hours at 23 C Given thatthe temperature of the lab in which the tests were made was set to 17 C and could not be brought upto 23 C it was decided to perform one test before the full strength was achieved and one after the fullstrength was achieved For practical reasons the former was performed 24 hours after gluing and thelatter 72 hours after gluing (one day longer than specified time of 48 hours in order to account forlower curing temperature)

The following plots show the relationship between crack opening size and strain reading forthe 3M DP100 adhesive Both of these tests were performed with strain gages with a gage factor of201 In the first test (Figure 10) the adhesive was allowed to set for 24 hours and in the second testfor 72 hours (Figure 11) Time-dependent tests were performed in order to evaluate the adhesivesrsquoperformance with time and to establish a minimum cure time for future lab tests

Figure 10 DP100 crack opening vs strain after 24 hours

Sensors 2018 18 1907 10 of 16

Figure 11 DP100 crack opening vs strain after 72 hours

The time-dependent tests for the 3M DP100 adhesive showed very interesting results The testsdemonstrated that the bond strength does show time-dependence The 24-hour DP100 test showedreadings up to a crack opening of about 015 mm whereas the 72-hour test showed readings up toa 026 mm crack The result is important because it verifies that the adhesive performs better whengiven a few days to cure

Similar tests were performed for the 3M DP105 adhesive The results are shown inFigures 12 and 13 below

Figure 12 DP105 crack opening vs strain after 24 hours

Sensors 2018 18 1907 11 of 16

Figure 13 DP105 crack opening vs strain after 72 hours

The DP105 adhesive continued to read strain for crack openings up to 045 mm compared to about026 mm for the DP100 adhesive Taking into account that all other properties of the two adhesivesare similar the DP105 adhesive is considered the more suitable adhesive for the installation of sensingsheets that are supposed to detect and survive structural damage

3 Application to a Sensing Sheet Prototype

31 Objective

Once the appropriate adhesive for the sensing sheet was selected its application was testedon a sensing sheet prototype Given that manufacturing of a sensing sheet at the prototype stageis expensive it was decided to perform only non-damage strain transfer tests so that the sensing sheetwould not be damaged and could be used for future research related to hardware

The first step in implementing the sensing sheet is to determine appropriate gage factors of theindividual sensors and observe how these gage factors change depending on the adhesive used tobond the sheet The same testing setup as described in Section 23 was used The sensing sheet wasinstalled in the third quarter of the beam ie 1282 mm from the left support The load was applied1069 mm from the left support

The prototype of the sensing sheet is composed of eight individual sensors each composed offour resistors forming a full bridge Each individual sensor was independently tested by bonding thesensing sheet to an aluminum beam then loading and unloading the beam with sand bottles similarto the strain-transfer test described in the previous section In total two sensing sheets were testedthe first glued with the stiff Araldite adhesive for reference purposes and the second glued with thesoft DP105 glue Each loading step lasted 15 s and values were sampled at 1000 Hz 10 seconds intoeach loading step The beam was loaded for seven loading steps no load B1 B1 and B2 B1 and B2and B3 B1 and B2 B1 and no load giving a total testing time of 105 seconds At each loading step thechange in output voltage was recorded Using the change in output voltage the input voltage and ananalytical strain model the gage factor was determined for each sensor under each loading conditionFor each sensor identical tests were performed three times yielding three gage factors Preliminarytests included air and beam surface temperature measurements but no significant temperature changeswere found during the test duration Figures 14 and 15 show the test setup

Sensors 2018 18 1907 12 of 16

Figure 14 Sensing sheet and individual sensor numbering convention

Figure 15 Gage factor test setup The sensing sheet is bonded to the top surface of the aluminum beamand is connected to the rigid printed circuit board The temperature reading device is shown in thebottom left corner

32 Results

For each loading step for each individual sensor apparent gage factors were determined underidentical loading conditions and they are shown in Figure 16 Note that these gage factors are apparentand not true as they combine true gage factor with strain transfer quality The figure shows thatconsistent results are obtained for all individual sensors within the same sheet In addition the gagefactors for sensors attached using the DP105 adhesive are about 25 times less than those for the sensorsattached using Araldite (see Table 1) This result is consistent with strain transfer test results describedearlier in Section 2

As expected smaller loads yielded greater gage factor errors than large loads due to lower signalto noise ratios This created dispersion in the results Figure 17 plots the gage factor uncertaintymeasured as standard deviation for each load step Indices 1 and 5 represent the beam loaded withbottle B1 2 and 4 represent loading with bottles B1 and B2 and 3 represents loading with all three

Sensors 2018 18 1907 13 of 16

bottles Loading step 3 has the smallest gage factor uncertainty with a standard deviation of about006 for 3M DP105 and 013 for Araldite Loading step 5 has the greatest gage factor uncertaintywith a standard deviation of about 033 for 3M DP105 and 054 for Araldite As Figure 17 showsthe Araldite and 3M DP105 have similar gage factor uncertainty trends except the Aralditersquos gagefactor uncertainties are consistently greater than those for 3M DP105 The reason for this could be theimperfections and connection issues in the flexible to rigid interface where a zero-insertion-force (ZIF)connector is used Unwanted impedance in this connection can vary by insertion and introduces asmall amount of noise into the differential voltage which is used to calculate strain This also showsthe need for an integrated system implementation on a flexible substrate and the need to reduce thenumber of interfaces to rigid boards as mentioned in previous works [18ndash20]

Figure 16 Apparent gage factors determined on individual sensors of sensing sheet prototypes gluedwith Araldite and 3M DP105 adhesives

Figure 17 Apparent gage factor uncertainties for various load steps

Sensors 2018 18 1907 14 of 16

Table 1 Apparent gage factor summary for Araldite and 3M DP105 adhesives (microlowast = meanσlowast = standard deviation)

Araldite GF 3M DP105 GF

microlowast 135 054σlowast 0240 0128

33 Discussion

The apparent gage factor for Araldite is greater than the apparent gage factor for 3M DP105adhesive because the Araldite is less flexible and transfers more strain from the structure to the sensorStrain is proportional to change in output voltage and inversely proportional to gage factor so if ananalytical strain model is used to determine gage factor a stiff adhesive will result in a larger changein output voltage and the gage factor will be higher Obtained results are consistent with the results ofthe strain transfer test presented in Section 2

For the sensing sheet bonded with Araldite sensors 2 and 7 showed high voltage output noise inthe change in output voltage values and therefore did not yield reasonable gage factor values For thesensing sheet bonded with 3M DP105 sensor 2 had high noise and did not yield reasonable gagefactor values Because change in output voltage values for sensor 2 were unreasonable in both theAraldite and 3M DP105 tests it is likely that there is a design or manufacturing flaw in the sensingsheet itself or a soldering problem on the rigid printed circuit board (see Figure 15) In the case ofsensor 7 in the Araldite-bonded sensor sheet the high noise can likely be attributed to a manufacturingflaw on the sensor sheet itself

Figure 16 shows that there is a range in apparent gage factor values (dispersion) from sensorto sensor One reason the gage factors might vary from sensor to sensor has to do with a possiblevoltage drop from the sensor to the rigid printed circuit board and reading unit Because the wiresin the sensing sheet have high resistance sensors that are located farthest from the reading unit(sensors 0 and 4 for example) may show a lower voltage output change than sensors that are closerto the reading unit (sensors 3 and 8) Furthermore a lower voltage output change will result in alower gage factor Figure 16 shows generally lower gage factors for sensors 0 and 4 validating thishypothesis Sensors 1 and 5 show higher gage factors and exhibit similar behavior Sensor 7 doesnot follow the trend described which may be attributed to a manufacturing flaw Another reasonfor varying gage factors between sensors is that the adhesive layerrsquos thickness might vary across thesensor sheet A thinner adhesive layer will permit more strain to be transferred to the sensor and thegage factor will be higher Conversely a thicker adhesive layer will transfer less strain to the sensorand will result in a lower computed gage factor

Finally Figure 17 and Table 1 show that the Araldite-bonded sensing sheet yields greater gagefactor uncertainties than the 3M DP105-bonded sensing sheet The Araldite adhesive is much moreviscous than the 3M DP105 adhesive and did not spread as easily when bonding the sensor sheet to thealuminum beam resulting in minor air pockets under the sensor sheet With the 3M DP105 adhesiveit was much easier to spread an even layer on the aluminum beam and no air pockets were visiblelikely increasing apparent gage factor consistency both between and within sensor sheets

4 Conclusions

The aim of this paper was to determine a suitable adhesive for installation of a novel sensing sheetonto steel structural members subjected to cracking Previous research has shown that for sensingsheets installed using a stiff adhesive (eg Araldite) the cracking in steel would induce immediatefailure of individual sensors of the sheet To address this challenge three soft 3M glues were consideredDP100 DP105 and DP125

The 3M DP125 adhesive was eliminated because the set time was too long to be useful inpractical applications Strain transfer tests were performed using strain gages to assess the quality

Sensors 2018 18 1907 15 of 16

of strain transfer of the two other glues It was shown that while the Araldite adhesive transfersa large percentage of strain to the sensors bonding with both the 3M DP100 and DP105 providessufficient sensitivity to detect small strain changes even in the non-damage range No conclusiveresults were found to justify the preference between the DP100 or DP105 from the non-damage studyso crack opening damage tests were performed 48-hour and 72-hour cure tests showed that the DP105adhesive was able to detect strain changes for larger crack opening sizes compared to the DP100 so itwas deemed more suitable for damage detection (cracking) in mechanically non-degrading materials(eg metals)

A sensing sheet prototype manufactured at Princeton University was then tested using the 3MDP105 adhesive For each of the working individual sensors apparent gage factors were determinedby loading and unloading a thin aluminum beam with sand bottles using theoretical strain valuesunder Euler-Bernoulli beam theory Uncertainties in the apparent gage factor values were found forboth Araldite and 3M DP105-bonded sensor sheets As expected the apparent gage factors for theAraldite adhesive were higher than for the 3M DP105 adhesive We know that an approximate gagefactor for bonding a sensor sheet with 3M DP105 adhesive would allow one to use the sensor sheet toidentify real strain values on structures

The consistency between the strain transfer tests performed on regular strain gages and the testsperformed on sensing sheets confirm that DP105 can be used for the installation of sensing sheets onsteel structural members for crack detection and characterization purposes

Author Contributions BG NV and JES created the concept of sensing sheet MG designed and carried outthe tests and data analysis he is the lead author of the paper CW and LEA developed and manufacturedsensing sheet prototype and associated hardware

Acknowledgments This research was in part supported by the Princeton Institute for the Science andTechnology of Materials (PRISM) and in part by the USDOT-RITA UTC Program grant no DTRT12-G- UTC16enabled through the Center for Advanced Infrastructure and Transportation (CAIT) at Rutgers UniversityThe authors would like to thank Y Hu L Huang N Lin W Rieutort-Louis J Sanz-Robinson T Liu S Wagnerand J Vocaturo from Princeton University for their precious help Kaitlyn Kliever helped improve the contents ofthis paper

Conflicts of Interest The authors declare no conflict of interest

References

1 Yao Y Glišic B Reliable damage detection and localization using direct strain sensing In Proceedings ofthe Sixth International IAMBAS Conference on Bridge Maintenance Safety and Management Stresa Italy8ndash12 July 2012 pp 714ndash721

2 Yao Y Tung S-TE Glišic B Crack detection and characterization techniquesmdashAn overview Struct ControlHealth Monit 2014 21 1387ndash1413 [CrossRef]

3 Tung S-T Yao Y Glišic B Sensing Sheet The Sensitivity of Thin-Film Full- Bridge Strain Sensors for CrackDetection and Characterization Meas Sci Technol 2014 25 14 [CrossRef]

4 Glišic B Chen J Hubbell D Streicker Bridge A comparison between Bragg-gratings long-gaugestrain and temperature sensors and Brillouin scattering-based distributed strain and temperature sensorsIn Proceedings of the SPIE Smart StructuresNDE and Materials + Nondestructive Evaluation and HealthMonitoring San Diego CA USA 6ndash10 March 2011

5 Loh KJ Hou TC Lynch JP Kotov NA Carbon Nanotube Sensing Skins for Spatial Strain and ImpactDamage Identification J Nondestruct Eval 2009 28 9ndash25 [CrossRef]

6 Pyo S Loh KJ Hou TC Jarva E Lynch JP A wireless impedance analyzer for automated tomographicmapping of a nanoengineered sensing skin Smart Struct Syst 2011 8 139ndash155 [CrossRef]

7 Pour-Ghaz M Weiss J Application of Frequency Selective Circuits for Crack Detection in ConcreteElements J ASTM Int 2011 8 1ndash11

8 Schumacher T Thostenson ET Development of structural carbon nanotube-based sensing composites forconcrete structures J Intell Mater Syst Struct 2014 25 1331ndash1339 [CrossRef]

Sensors 2018 18 1907 4 of 16

It is important to note that while the general sensing principle of sensing sheet is based onstrain measurement its purpose is not to accurately measure strain magnitudes but rather to detectlocalize and quantify unusual behaviors that manifest in the form of strain-field anomalies (egcracking of concrete fatigue cracking and bowing of steel) Being in direct contact with sensing sheetindividual sensors the strain-field anomalies create unusually high strain changes in these sensorscompared with individual sensors that are not in direct contact with the anomaly This unusually highstrain change is used to directly detect and localize the anomaly The extent of the anomaly is theninferred from geographical locations of affected individual sensors This principle is shown in Figure 3

The applications of sensing sheet could be on a variety of civil and mechanical structures andinfrastructure including but not limited to bridges buildings tunnels pipelines wind turbinesaircrafts etc Given that accurate measurement of strain is not required the imperfections andphenomena that affect measurements in the range of usual strain changes (eg potential rheologicalstrain in glue lower strain transfer due to gluing over the paint of original structure etc) wouldnot significantly affect the performance of the sensing sheet for this function More detail is given inliterature [1ndash316ndash20]

Figure 3 Schematic representation of the sensing sheet based on LAE and ICs [3]

Yao and Glišic [16] performed preliminary tests to evaluate whether an early-stage LAE prototypehas the potential to detect damage such as a crack [1] The authors tested a full-bridge resistance-basedstrain sensor as a preliminary step towards LAE

Series of tests were performed at the Carleton Laboratory at Columbia University on an LAE-basedstrain sensing sheet prototype in order to evaluate the sensorrsquos ability to detect cracks in steel platespecimens [16] In these tests the sensors were bonded to the steel samples with a notch and wereexposed to fatigue cycling so that the crack occurrence and propagation could be predicted While thetests were successful from the point of view of damage detection they revealed a challenge individualsensors would fail immediately upon damage occurrence thereby limiting the ability of sensing sheetsto follow and quantify damage progression That challenge is addressed in this paper

Sensors 2018 18 1907 5 of 16

2 Selecting a Suitable Adhesive

21 Objective

Because manufacturing sensing sheets at the prototype stage is expensive the following tests areperformed using existing identical strain-gages

A critical component of a sensorrsquos functionality is the way the sensor is fixed to the structurersquossurface Because the sensing sheets involved in this study are formed on a polyimide substratean adhesive is the best fixing option At minimum a suitable adhesive should maintain bond betweenthe sensor and structure under various loads damage types and environmental conditions An equallyimportant consideration involves the adhesiversquos elastic behavior particularly in materials that do notmechanically degrade when damaged Due to the heterogeneous nature and particle flaws concretedegrades when subjected to tensile stresses When subjecting a strain sensor bonded to concreteto moderate strains the bond between the sensor and concrete is stronger than the bond betweenaggregate particles in the concrete causing the concrete to break apart near the crack In this case thestrain in the sensor is defined over the strain length shown in Figure 4 Conversely when subjectinga strain sensor bonded to non-degrading materials such as aluminum and steel the bond betweenparticles in the material is much stronger than the bond between the sensing sheet and structure sothe strain is defined over the sample separation (See Figure 5) leading to much higher strain valuesMechanical strain in the sensor is calculated as the change in length divided by original length Fora perfectly non-degrading material that has no initial separation any separation will theoreticallycause infinite strain because original separation length is zero Hence in non-degrading materials thesensor will fail soon after crack occurrence which in turn will disable monitoring of crack size and itsevolution Therefore the purpose of this study is to identify and both qualitatively and quantitativelyevaluate an adhesive to be used for two-dimensional sensing on non-degrading materials so that thesensor (sensing sheet) survives crack occurrence and continues monitoring the crack size

The soft adhesives considered in this research are expected to have viscoelastic behaviorand consequently rheological effects in adhesives may influence values of strain measurements overtime However as the main aim of this research was to identify an adhesive that enables the sensingsheet to survive the damage to the structure only short-term effects were analyzed and long-termeffects will be a subject of future work This approach was considered suitable as the maximum stressin sensing sheet happens instantaneously with occurrence of damage (rheological effects are actuallybeneficial as they bring relaxation) In addition rheological effects are small compared to strain changesdue to damage and thus they do not significantly affect its damage detection capability (they affectstrain measurement but accurate strain measurement is not the purpose of the sensing sheet)

Figure 4 Strain length vs sample separation on degrading materials Degradation is indicated byred lines

Sensors 2018 18 1907 6 of 16

Figure 5 Strain length vs sample separation on non-degrading materials

22 Qualitative Tests

Qualitative observations were performed for three high-strength flexible adhesive alternatives3M DP100 DP105 and DP125 Initially these three adhesives were tested on paper samples in orderto gain an understanding of viscosity flexibility and set time It was found that the three adhesiveslisted were very flexible and could be bent and twisted (along with the substrate paper sheet) byhand without much effort A sample of a flexible adhesive applied to paper and bent can be seenin Figure 6 below

Figure 6 Sample of a flexible adhesive applied to paper

The 3M DP100 had a set time of about five minutes the DP105 had a set time of about 20 minutesand the DP125 had a set time which exceeded half an hour An ideal set time is neither too short(to give adequate time to place the sensor sheet) nor too long (to avoid sensor displacement caused byaccidental contact shortly after placement) A set time exceeding a half hour is undesirable because itincreases the risk of accidentally moving or removing the sensor during the installation phase Becausethe DP125 adhesive set time exceeded half an hour it was eliminated and the DP100 and DP105adhesives were further tested

Sensors 2018 18 1907 7 of 16

23 Quantitative Tests Loading and Unloading

Strain tests were performed in order to understand how much strain is transferred to the sensorfrom a monitored material given that the adhesives are soft and some loss of strain transfer is expectedThe test involved bonding three sensors to the top of a thin aluminum simply supported beam andloading and unloading the beam with known point forces Due to limitations of available testingset-up bottles filled with sand were used to load the beam The beam was simply supported at bothends as shown in Figure 7 It had span of 1710 mm thickness of 913 mm and width of 254 mmThe beam was made of aluminum and had a Youngrsquos modulus of 68 GPa Strain gages were placed at855 mm from the left support (in the middle of the beam) The sand bottles B1 B2 and B3 had masses5942 g (58 kN) 5981 g (59 kN) and 5977 g (59 kN) respectively For each load step the strain valuewas recorded with an Omega DP41-S reading unit (Omega Engineering Stamford CT USA) The totalloading and unloading time was less than 2 minutes The three tested adhesives were a stiff Araldite2012 adhesive (used by Yao and Glišic in previous studies [16]) 3M DP100 and 3M DP105 The stiffadhesive was used for reference purposes as it is assumed to have high quantity of strain transferdue to its high stiffness The aluminum beam and three bonded sensors used in the test are shown inFigure 7 below

Figure 7 Strain transfer test set-up (a) view of beam without sensors (b) support detail (c) view ofsensors installed using Araldite 2012 3M DP 100 and 3M DP 105 (d) sensor details

The results of the strain transfer tests are shown in Figure 8 An analytic strain model wasdeveloped for the aluminum beam The model was based on the linear theory of beams and ispresented in Equation (5)

εs =M times ys

EI(5)

where M = FL24mdashbending moment at location of the sensors (F is force applied using bottles placedin the middle of the beam L = 1710 mm is length of the beam)

E = 68 GPamdashYoungrsquos modulus of aluminumI = bh312mdashmoment of inertia of the beamrsquos cross-section (b = 254 mm h = 913 mm)

and ys = h2 + 02 mmmdashdistance of sensors from the horizontal principal axis of the cross-section(where 02 mm accounts for estimated thickness of the glue plus half the thickness of the strain gage)

Sensors 2018 18 1907 8 of 16

Figure 8 shows the instantaneous bending strain for seven load steps The first and last load stepscorrespond with no beam loading the second and sixth load steps correspond with the beam loadedwith sand bottle B1 the third and fifth load steps correspond with the beam loaded with sand bottlesB1 and B2 and the fourth load step corresponds with the beam loaded with all three sand bottlesAs expected the stiff Araldite adhesive consistently transfers the greatest fraction of strain from thestructure to the sensor and none of the adhesives transferred all strain to the sensor The strain valuesmeasured by sensors glued with Araldite were about 25 times greater than those measured by theother two glues Figure 8 shows that while DP100 transfers slightly more strain than DP105 in generalthere is no significant strain transfer difference between the 3M DP100 and 3M DP105 adhesivesBoth adhesives ldquoloserdquo more than 50 percent of strain however this is acceptable as the main purposeof the sensing sheet is damage detection and not strain monitoring Moreover linear response ofboth adhesives indicates that lack in strain transfer can be accounted for through calibration and gagefactor adjustments This is effectively done in Section 3 of this paper where strain transfer was testedon sensing sheet prototypes

Figure 8 Strain transfer results for various adhesives

24 Damage Tests Crack Opening vs Strain Test

The aim of the damage test was to assess the behavior of adhesives at crack locations andto see how it affects the survival of the sensors For these tests a sensor was bonded to ldquoclosedrdquoaluminum plates and the adhesive was given time to cure Once the adhesive reached suitablestrength an artificial crack was generated by pulling the aluminum plates apart thereby creating a gapbetween them The test setup is shown in Figure 9 Each simulated crack opening increment involvedpulling the aluminum plates apart by 0001 inches taking the strain reading and then waiting fiveminutes to record the ldquorelaxedrdquo strain reading Relaxation is expected due to creep in sensor polyimidesubstrate as well as due to creep in the adhesive exposed to high stress levels This process wasrepeated until failure (ie until the sensor could no longer read the strain due to rupture or the strainvalues began to decrease due to debonding or the adhesive breaking) The tests were performed withVishay 3800 strain reading units and the gage factor was calibrated to the manufacturerrsquos specificationsAn example of post-test damage is shown below in Figure 9

Sensors 2018 18 1907 9 of 16

Figure 9 Crack test setup

Based on the manufacturerrsquos specifications both adhesives achieve 80 of their full strengths ifcured for 20 minutes at 23 C and they achieve full strength if cured for 48 hours at 23 C Given thatthe temperature of the lab in which the tests were made was set to 17 C and could not be brought upto 23 C it was decided to perform one test before the full strength was achieved and one after the fullstrength was achieved For practical reasons the former was performed 24 hours after gluing and thelatter 72 hours after gluing (one day longer than specified time of 48 hours in order to account forlower curing temperature)

The following plots show the relationship between crack opening size and strain reading forthe 3M DP100 adhesive Both of these tests were performed with strain gages with a gage factor of201 In the first test (Figure 10) the adhesive was allowed to set for 24 hours and in the second testfor 72 hours (Figure 11) Time-dependent tests were performed in order to evaluate the adhesivesrsquoperformance with time and to establish a minimum cure time for future lab tests

Figure 10 DP100 crack opening vs strain after 24 hours

Sensors 2018 18 1907 10 of 16

Figure 11 DP100 crack opening vs strain after 72 hours

The time-dependent tests for the 3M DP100 adhesive showed very interesting results The testsdemonstrated that the bond strength does show time-dependence The 24-hour DP100 test showedreadings up to a crack opening of about 015 mm whereas the 72-hour test showed readings up toa 026 mm crack The result is important because it verifies that the adhesive performs better whengiven a few days to cure

Similar tests were performed for the 3M DP105 adhesive The results are shown inFigures 12 and 13 below

Figure 12 DP105 crack opening vs strain after 24 hours

Sensors 2018 18 1907 11 of 16

Figure 13 DP105 crack opening vs strain after 72 hours

The DP105 adhesive continued to read strain for crack openings up to 045 mm compared to about026 mm for the DP100 adhesive Taking into account that all other properties of the two adhesivesare similar the DP105 adhesive is considered the more suitable adhesive for the installation of sensingsheets that are supposed to detect and survive structural damage

3 Application to a Sensing Sheet Prototype

31 Objective

Once the appropriate adhesive for the sensing sheet was selected its application was testedon a sensing sheet prototype Given that manufacturing of a sensing sheet at the prototype stageis expensive it was decided to perform only non-damage strain transfer tests so that the sensing sheetwould not be damaged and could be used for future research related to hardware

The first step in implementing the sensing sheet is to determine appropriate gage factors of theindividual sensors and observe how these gage factors change depending on the adhesive used tobond the sheet The same testing setup as described in Section 23 was used The sensing sheet wasinstalled in the third quarter of the beam ie 1282 mm from the left support The load was applied1069 mm from the left support

The prototype of the sensing sheet is composed of eight individual sensors each composed offour resistors forming a full bridge Each individual sensor was independently tested by bonding thesensing sheet to an aluminum beam then loading and unloading the beam with sand bottles similarto the strain-transfer test described in the previous section In total two sensing sheets were testedthe first glued with the stiff Araldite adhesive for reference purposes and the second glued with thesoft DP105 glue Each loading step lasted 15 s and values were sampled at 1000 Hz 10 seconds intoeach loading step The beam was loaded for seven loading steps no load B1 B1 and B2 B1 and B2and B3 B1 and B2 B1 and no load giving a total testing time of 105 seconds At each loading step thechange in output voltage was recorded Using the change in output voltage the input voltage and ananalytical strain model the gage factor was determined for each sensor under each loading conditionFor each sensor identical tests were performed three times yielding three gage factors Preliminarytests included air and beam surface temperature measurements but no significant temperature changeswere found during the test duration Figures 14 and 15 show the test setup

Sensors 2018 18 1907 12 of 16

Figure 14 Sensing sheet and individual sensor numbering convention

Figure 15 Gage factor test setup The sensing sheet is bonded to the top surface of the aluminum beamand is connected to the rigid printed circuit board The temperature reading device is shown in thebottom left corner

32 Results

For each loading step for each individual sensor apparent gage factors were determined underidentical loading conditions and they are shown in Figure 16 Note that these gage factors are apparentand not true as they combine true gage factor with strain transfer quality The figure shows thatconsistent results are obtained for all individual sensors within the same sheet In addition the gagefactors for sensors attached using the DP105 adhesive are about 25 times less than those for the sensorsattached using Araldite (see Table 1) This result is consistent with strain transfer test results describedearlier in Section 2

As expected smaller loads yielded greater gage factor errors than large loads due to lower signalto noise ratios This created dispersion in the results Figure 17 plots the gage factor uncertaintymeasured as standard deviation for each load step Indices 1 and 5 represent the beam loaded withbottle B1 2 and 4 represent loading with bottles B1 and B2 and 3 represents loading with all three

Sensors 2018 18 1907 13 of 16

bottles Loading step 3 has the smallest gage factor uncertainty with a standard deviation of about006 for 3M DP105 and 013 for Araldite Loading step 5 has the greatest gage factor uncertaintywith a standard deviation of about 033 for 3M DP105 and 054 for Araldite As Figure 17 showsthe Araldite and 3M DP105 have similar gage factor uncertainty trends except the Aralditersquos gagefactor uncertainties are consistently greater than those for 3M DP105 The reason for this could be theimperfections and connection issues in the flexible to rigid interface where a zero-insertion-force (ZIF)connector is used Unwanted impedance in this connection can vary by insertion and introduces asmall amount of noise into the differential voltage which is used to calculate strain This also showsthe need for an integrated system implementation on a flexible substrate and the need to reduce thenumber of interfaces to rigid boards as mentioned in previous works [18ndash20]

Figure 16 Apparent gage factors determined on individual sensors of sensing sheet prototypes gluedwith Araldite and 3M DP105 adhesives

Figure 17 Apparent gage factor uncertainties for various load steps

Sensors 2018 18 1907 14 of 16

Table 1 Apparent gage factor summary for Araldite and 3M DP105 adhesives (microlowast = meanσlowast = standard deviation)

Araldite GF 3M DP105 GF

microlowast 135 054σlowast 0240 0128

33 Discussion

The apparent gage factor for Araldite is greater than the apparent gage factor for 3M DP105adhesive because the Araldite is less flexible and transfers more strain from the structure to the sensorStrain is proportional to change in output voltage and inversely proportional to gage factor so if ananalytical strain model is used to determine gage factor a stiff adhesive will result in a larger changein output voltage and the gage factor will be higher Obtained results are consistent with the results ofthe strain transfer test presented in Section 2

For the sensing sheet bonded with Araldite sensors 2 and 7 showed high voltage output noise inthe change in output voltage values and therefore did not yield reasonable gage factor values For thesensing sheet bonded with 3M DP105 sensor 2 had high noise and did not yield reasonable gagefactor values Because change in output voltage values for sensor 2 were unreasonable in both theAraldite and 3M DP105 tests it is likely that there is a design or manufacturing flaw in the sensingsheet itself or a soldering problem on the rigid printed circuit board (see Figure 15) In the case ofsensor 7 in the Araldite-bonded sensor sheet the high noise can likely be attributed to a manufacturingflaw on the sensor sheet itself

Figure 16 shows that there is a range in apparent gage factor values (dispersion) from sensorto sensor One reason the gage factors might vary from sensor to sensor has to do with a possiblevoltage drop from the sensor to the rigid printed circuit board and reading unit Because the wiresin the sensing sheet have high resistance sensors that are located farthest from the reading unit(sensors 0 and 4 for example) may show a lower voltage output change than sensors that are closerto the reading unit (sensors 3 and 8) Furthermore a lower voltage output change will result in alower gage factor Figure 16 shows generally lower gage factors for sensors 0 and 4 validating thishypothesis Sensors 1 and 5 show higher gage factors and exhibit similar behavior Sensor 7 doesnot follow the trend described which may be attributed to a manufacturing flaw Another reasonfor varying gage factors between sensors is that the adhesive layerrsquos thickness might vary across thesensor sheet A thinner adhesive layer will permit more strain to be transferred to the sensor and thegage factor will be higher Conversely a thicker adhesive layer will transfer less strain to the sensorand will result in a lower computed gage factor

Finally Figure 17 and Table 1 show that the Araldite-bonded sensing sheet yields greater gagefactor uncertainties than the 3M DP105-bonded sensing sheet The Araldite adhesive is much moreviscous than the 3M DP105 adhesive and did not spread as easily when bonding the sensor sheet to thealuminum beam resulting in minor air pockets under the sensor sheet With the 3M DP105 adhesiveit was much easier to spread an even layer on the aluminum beam and no air pockets were visiblelikely increasing apparent gage factor consistency both between and within sensor sheets

4 Conclusions

The aim of this paper was to determine a suitable adhesive for installation of a novel sensing sheetonto steel structural members subjected to cracking Previous research has shown that for sensingsheets installed using a stiff adhesive (eg Araldite) the cracking in steel would induce immediatefailure of individual sensors of the sheet To address this challenge three soft 3M glues were consideredDP100 DP105 and DP125

The 3M DP125 adhesive was eliminated because the set time was too long to be useful inpractical applications Strain transfer tests were performed using strain gages to assess the quality

Sensors 2018 18 1907 15 of 16

of strain transfer of the two other glues It was shown that while the Araldite adhesive transfersa large percentage of strain to the sensors bonding with both the 3M DP100 and DP105 providessufficient sensitivity to detect small strain changes even in the non-damage range No conclusiveresults were found to justify the preference between the DP100 or DP105 from the non-damage studyso crack opening damage tests were performed 48-hour and 72-hour cure tests showed that the DP105adhesive was able to detect strain changes for larger crack opening sizes compared to the DP100 so itwas deemed more suitable for damage detection (cracking) in mechanically non-degrading materials(eg metals)

A sensing sheet prototype manufactured at Princeton University was then tested using the 3MDP105 adhesive For each of the working individual sensors apparent gage factors were determinedby loading and unloading a thin aluminum beam with sand bottles using theoretical strain valuesunder Euler-Bernoulli beam theory Uncertainties in the apparent gage factor values were found forboth Araldite and 3M DP105-bonded sensor sheets As expected the apparent gage factors for theAraldite adhesive were higher than for the 3M DP105 adhesive We know that an approximate gagefactor for bonding a sensor sheet with 3M DP105 adhesive would allow one to use the sensor sheet toidentify real strain values on structures

The consistency between the strain transfer tests performed on regular strain gages and the testsperformed on sensing sheets confirm that DP105 can be used for the installation of sensing sheets onsteel structural members for crack detection and characterization purposes

Author Contributions BG NV and JES created the concept of sensing sheet MG designed and carried outthe tests and data analysis he is the lead author of the paper CW and LEA developed and manufacturedsensing sheet prototype and associated hardware

Acknowledgments This research was in part supported by the Princeton Institute for the Science andTechnology of Materials (PRISM) and in part by the USDOT-RITA UTC Program grant no DTRT12-G- UTC16enabled through the Center for Advanced Infrastructure and Transportation (CAIT) at Rutgers UniversityThe authors would like to thank Y Hu L Huang N Lin W Rieutort-Louis J Sanz-Robinson T Liu S Wagnerand J Vocaturo from Princeton University for their precious help Kaitlyn Kliever helped improve the contents ofthis paper

Conflicts of Interest The authors declare no conflict of interest

References

1 Yao Y Glišic B Reliable damage detection and localization using direct strain sensing In Proceedings ofthe Sixth International IAMBAS Conference on Bridge Maintenance Safety and Management Stresa Italy8ndash12 July 2012 pp 714ndash721

2 Yao Y Tung S-TE Glišic B Crack detection and characterization techniquesmdashAn overview Struct ControlHealth Monit 2014 21 1387ndash1413 [CrossRef]

3 Tung S-T Yao Y Glišic B Sensing Sheet The Sensitivity of Thin-Film Full- Bridge Strain Sensors for CrackDetection and Characterization Meas Sci Technol 2014 25 14 [CrossRef]

4 Glišic B Chen J Hubbell D Streicker Bridge A comparison between Bragg-gratings long-gaugestrain and temperature sensors and Brillouin scattering-based distributed strain and temperature sensorsIn Proceedings of the SPIE Smart StructuresNDE and Materials + Nondestructive Evaluation and HealthMonitoring San Diego CA USA 6ndash10 March 2011

5 Loh KJ Hou TC Lynch JP Kotov NA Carbon Nanotube Sensing Skins for Spatial Strain and ImpactDamage Identification J Nondestruct Eval 2009 28 9ndash25 [CrossRef]

6 Pyo S Loh KJ Hou TC Jarva E Lynch JP A wireless impedance analyzer for automated tomographicmapping of a nanoengineered sensing skin Smart Struct Syst 2011 8 139ndash155 [CrossRef]

7 Pour-Ghaz M Weiss J Application of Frequency Selective Circuits for Crack Detection in ConcreteElements J ASTM Int 2011 8 1ndash11

8 Schumacher T Thostenson ET Development of structural carbon nanotube-based sensing composites forconcrete structures J Intell Mater Syst Struct 2014 25 1331ndash1339 [CrossRef]

Sensors 2018 18 1907 5 of 16

2 Selecting a Suitable Adhesive

21 Objective

Because manufacturing sensing sheets at the prototype stage is expensive the following tests areperformed using existing identical strain-gages

A critical component of a sensorrsquos functionality is the way the sensor is fixed to the structurersquossurface Because the sensing sheets involved in this study are formed on a polyimide substratean adhesive is the best fixing option At minimum a suitable adhesive should maintain bond betweenthe sensor and structure under various loads damage types and environmental conditions An equallyimportant consideration involves the adhesiversquos elastic behavior particularly in materials that do notmechanically degrade when damaged Due to the heterogeneous nature and particle flaws concretedegrades when subjected to tensile stresses When subjecting a strain sensor bonded to concreteto moderate strains the bond between the sensor and concrete is stronger than the bond betweenaggregate particles in the concrete causing the concrete to break apart near the crack In this case thestrain in the sensor is defined over the strain length shown in Figure 4 Conversely when subjectinga strain sensor bonded to non-degrading materials such as aluminum and steel the bond betweenparticles in the material is much stronger than the bond between the sensing sheet and structure sothe strain is defined over the sample separation (See Figure 5) leading to much higher strain valuesMechanical strain in the sensor is calculated as the change in length divided by original length Fora perfectly non-degrading material that has no initial separation any separation will theoreticallycause infinite strain because original separation length is zero Hence in non-degrading materials thesensor will fail soon after crack occurrence which in turn will disable monitoring of crack size and itsevolution Therefore the purpose of this study is to identify and both qualitatively and quantitativelyevaluate an adhesive to be used for two-dimensional sensing on non-degrading materials so that thesensor (sensing sheet) survives crack occurrence and continues monitoring the crack size

The soft adhesives considered in this research are expected to have viscoelastic behaviorand consequently rheological effects in adhesives may influence values of strain measurements overtime However as the main aim of this research was to identify an adhesive that enables the sensingsheet to survive the damage to the structure only short-term effects were analyzed and long-termeffects will be a subject of future work This approach was considered suitable as the maximum stressin sensing sheet happens instantaneously with occurrence of damage (rheological effects are actuallybeneficial as they bring relaxation) In addition rheological effects are small compared to strain changesdue to damage and thus they do not significantly affect its damage detection capability (they affectstrain measurement but accurate strain measurement is not the purpose of the sensing sheet)

Figure 4 Strain length vs sample separation on degrading materials Degradation is indicated byred lines

Sensors 2018 18 1907 6 of 16

Figure 5 Strain length vs sample separation on non-degrading materials

22 Qualitative Tests

Qualitative observations were performed for three high-strength flexible adhesive alternatives3M DP100 DP105 and DP125 Initially these three adhesives were tested on paper samples in orderto gain an understanding of viscosity flexibility and set time It was found that the three adhesiveslisted were very flexible and could be bent and twisted (along with the substrate paper sheet) byhand without much effort A sample of a flexible adhesive applied to paper and bent can be seenin Figure 6 below

Figure 6 Sample of a flexible adhesive applied to paper

The 3M DP100 had a set time of about five minutes the DP105 had a set time of about 20 minutesand the DP125 had a set time which exceeded half an hour An ideal set time is neither too short(to give adequate time to place the sensor sheet) nor too long (to avoid sensor displacement caused byaccidental contact shortly after placement) A set time exceeding a half hour is undesirable because itincreases the risk of accidentally moving or removing the sensor during the installation phase Becausethe DP125 adhesive set time exceeded half an hour it was eliminated and the DP100 and DP105adhesives were further tested

Sensors 2018 18 1907 7 of 16

23 Quantitative Tests Loading and Unloading

Strain tests were performed in order to understand how much strain is transferred to the sensorfrom a monitored material given that the adhesives are soft and some loss of strain transfer is expectedThe test involved bonding three sensors to the top of a thin aluminum simply supported beam andloading and unloading the beam with known point forces Due to limitations of available testingset-up bottles filled with sand were used to load the beam The beam was simply supported at bothends as shown in Figure 7 It had span of 1710 mm thickness of 913 mm and width of 254 mmThe beam was made of aluminum and had a Youngrsquos modulus of 68 GPa Strain gages were placed at855 mm from the left support (in the middle of the beam) The sand bottles B1 B2 and B3 had masses5942 g (58 kN) 5981 g (59 kN) and 5977 g (59 kN) respectively For each load step the strain valuewas recorded with an Omega DP41-S reading unit (Omega Engineering Stamford CT USA) The totalloading and unloading time was less than 2 minutes The three tested adhesives were a stiff Araldite2012 adhesive (used by Yao and Glišic in previous studies [16]) 3M DP100 and 3M DP105 The stiffadhesive was used for reference purposes as it is assumed to have high quantity of strain transferdue to its high stiffness The aluminum beam and three bonded sensors used in the test are shown inFigure 7 below

Figure 7 Strain transfer test set-up (a) view of beam without sensors (b) support detail (c) view ofsensors installed using Araldite 2012 3M DP 100 and 3M DP 105 (d) sensor details

The results of the strain transfer tests are shown in Figure 8 An analytic strain model wasdeveloped for the aluminum beam The model was based on the linear theory of beams and ispresented in Equation (5)

εs =M times ys

EI(5)

where M = FL24mdashbending moment at location of the sensors (F is force applied using bottles placedin the middle of the beam L = 1710 mm is length of the beam)

E = 68 GPamdashYoungrsquos modulus of aluminumI = bh312mdashmoment of inertia of the beamrsquos cross-section (b = 254 mm h = 913 mm)

and ys = h2 + 02 mmmdashdistance of sensors from the horizontal principal axis of the cross-section(where 02 mm accounts for estimated thickness of the glue plus half the thickness of the strain gage)

Sensors 2018 18 1907 8 of 16

Figure 8 shows the instantaneous bending strain for seven load steps The first and last load stepscorrespond with no beam loading the second and sixth load steps correspond with the beam loadedwith sand bottle B1 the third and fifth load steps correspond with the beam loaded with sand bottlesB1 and B2 and the fourth load step corresponds with the beam loaded with all three sand bottlesAs expected the stiff Araldite adhesive consistently transfers the greatest fraction of strain from thestructure to the sensor and none of the adhesives transferred all strain to the sensor The strain valuesmeasured by sensors glued with Araldite were about 25 times greater than those measured by theother two glues Figure 8 shows that while DP100 transfers slightly more strain than DP105 in generalthere is no significant strain transfer difference between the 3M DP100 and 3M DP105 adhesivesBoth adhesives ldquoloserdquo more than 50 percent of strain however this is acceptable as the main purposeof the sensing sheet is damage detection and not strain monitoring Moreover linear response ofboth adhesives indicates that lack in strain transfer can be accounted for through calibration and gagefactor adjustments This is effectively done in Section 3 of this paper where strain transfer was testedon sensing sheet prototypes

Figure 8 Strain transfer results for various adhesives

24 Damage Tests Crack Opening vs Strain Test

The aim of the damage test was to assess the behavior of adhesives at crack locations andto see how it affects the survival of the sensors For these tests a sensor was bonded to ldquoclosedrdquoaluminum plates and the adhesive was given time to cure Once the adhesive reached suitablestrength an artificial crack was generated by pulling the aluminum plates apart thereby creating a gapbetween them The test setup is shown in Figure 9 Each simulated crack opening increment involvedpulling the aluminum plates apart by 0001 inches taking the strain reading and then waiting fiveminutes to record the ldquorelaxedrdquo strain reading Relaxation is expected due to creep in sensor polyimidesubstrate as well as due to creep in the adhesive exposed to high stress levels This process wasrepeated until failure (ie until the sensor could no longer read the strain due to rupture or the strainvalues began to decrease due to debonding or the adhesive breaking) The tests were performed withVishay 3800 strain reading units and the gage factor was calibrated to the manufacturerrsquos specificationsAn example of post-test damage is shown below in Figure 9

Sensors 2018 18 1907 9 of 16

Figure 9 Crack test setup

Based on the manufacturerrsquos specifications both adhesives achieve 80 of their full strengths ifcured for 20 minutes at 23 C and they achieve full strength if cured for 48 hours at 23 C Given thatthe temperature of the lab in which the tests were made was set to 17 C and could not be brought upto 23 C it was decided to perform one test before the full strength was achieved and one after the fullstrength was achieved For practical reasons the former was performed 24 hours after gluing and thelatter 72 hours after gluing (one day longer than specified time of 48 hours in order to account forlower curing temperature)

The following plots show the relationship between crack opening size and strain reading forthe 3M DP100 adhesive Both of these tests were performed with strain gages with a gage factor of201 In the first test (Figure 10) the adhesive was allowed to set for 24 hours and in the second testfor 72 hours (Figure 11) Time-dependent tests were performed in order to evaluate the adhesivesrsquoperformance with time and to establish a minimum cure time for future lab tests

Figure 10 DP100 crack opening vs strain after 24 hours

Sensors 2018 18 1907 10 of 16

Figure 11 DP100 crack opening vs strain after 72 hours

The time-dependent tests for the 3M DP100 adhesive showed very interesting results The testsdemonstrated that the bond strength does show time-dependence The 24-hour DP100 test showedreadings up to a crack opening of about 015 mm whereas the 72-hour test showed readings up toa 026 mm crack The result is important because it verifies that the adhesive performs better whengiven a few days to cure

Similar tests were performed for the 3M DP105 adhesive The results are shown inFigures 12 and 13 below

Figure 12 DP105 crack opening vs strain after 24 hours

Sensors 2018 18 1907 11 of 16

Figure 13 DP105 crack opening vs strain after 72 hours

The DP105 adhesive continued to read strain for crack openings up to 045 mm compared to about026 mm for the DP100 adhesive Taking into account that all other properties of the two adhesivesare similar the DP105 adhesive is considered the more suitable adhesive for the installation of sensingsheets that are supposed to detect and survive structural damage

3 Application to a Sensing Sheet Prototype

31 Objective

Once the appropriate adhesive for the sensing sheet was selected its application was testedon a sensing sheet prototype Given that manufacturing of a sensing sheet at the prototype stageis expensive it was decided to perform only non-damage strain transfer tests so that the sensing sheetwould not be damaged and could be used for future research related to hardware

The first step in implementing the sensing sheet is to determine appropriate gage factors of theindividual sensors and observe how these gage factors change depending on the adhesive used tobond the sheet The same testing setup as described in Section 23 was used The sensing sheet wasinstalled in the third quarter of the beam ie 1282 mm from the left support The load was applied1069 mm from the left support

The prototype of the sensing sheet is composed of eight individual sensors each composed offour resistors forming a full bridge Each individual sensor was independently tested by bonding thesensing sheet to an aluminum beam then loading and unloading the beam with sand bottles similarto the strain-transfer test described in the previous section In total two sensing sheets were testedthe first glued with the stiff Araldite adhesive for reference purposes and the second glued with thesoft DP105 glue Each loading step lasted 15 s and values were sampled at 1000 Hz 10 seconds intoeach loading step The beam was loaded for seven loading steps no load B1 B1 and B2 B1 and B2and B3 B1 and B2 B1 and no load giving a total testing time of 105 seconds At each loading step thechange in output voltage was recorded Using the change in output voltage the input voltage and ananalytical strain model the gage factor was determined for each sensor under each loading conditionFor each sensor identical tests were performed three times yielding three gage factors Preliminarytests included air and beam surface temperature measurements but no significant temperature changeswere found during the test duration Figures 14 and 15 show the test setup

Sensors 2018 18 1907 12 of 16

Figure 14 Sensing sheet and individual sensor numbering convention

Figure 15 Gage factor test setup The sensing sheet is bonded to the top surface of the aluminum beamand is connected to the rigid printed circuit board The temperature reading device is shown in thebottom left corner

32 Results

For each loading step for each individual sensor apparent gage factors were determined underidentical loading conditions and they are shown in Figure 16 Note that these gage factors are apparentand not true as they combine true gage factor with strain transfer quality The figure shows thatconsistent results are obtained for all individual sensors within the same sheet In addition the gagefactors for sensors attached using the DP105 adhesive are about 25 times less than those for the sensorsattached using Araldite (see Table 1) This result is consistent with strain transfer test results describedearlier in Section 2

As expected smaller loads yielded greater gage factor errors than large loads due to lower signalto noise ratios This created dispersion in the results Figure 17 plots the gage factor uncertaintymeasured as standard deviation for each load step Indices 1 and 5 represent the beam loaded withbottle B1 2 and 4 represent loading with bottles B1 and B2 and 3 represents loading with all three

Sensors 2018 18 1907 13 of 16

bottles Loading step 3 has the smallest gage factor uncertainty with a standard deviation of about006 for 3M DP105 and 013 for Araldite Loading step 5 has the greatest gage factor uncertaintywith a standard deviation of about 033 for 3M DP105 and 054 for Araldite As Figure 17 showsthe Araldite and 3M DP105 have similar gage factor uncertainty trends except the Aralditersquos gagefactor uncertainties are consistently greater than those for 3M DP105 The reason for this could be theimperfections and connection issues in the flexible to rigid interface where a zero-insertion-force (ZIF)connector is used Unwanted impedance in this connection can vary by insertion and introduces asmall amount of noise into the differential voltage which is used to calculate strain This also showsthe need for an integrated system implementation on a flexible substrate and the need to reduce thenumber of interfaces to rigid boards as mentioned in previous works [18ndash20]

Figure 16 Apparent gage factors determined on individual sensors of sensing sheet prototypes gluedwith Araldite and 3M DP105 adhesives

Figure 17 Apparent gage factor uncertainties for various load steps

Sensors 2018 18 1907 14 of 16

Table 1 Apparent gage factor summary for Araldite and 3M DP105 adhesives (microlowast = meanσlowast = standard deviation)

Araldite GF 3M DP105 GF

microlowast 135 054σlowast 0240 0128

33 Discussion

The apparent gage factor for Araldite is greater than the apparent gage factor for 3M DP105adhesive because the Araldite is less flexible and transfers more strain from the structure to the sensorStrain is proportional to change in output voltage and inversely proportional to gage factor so if ananalytical strain model is used to determine gage factor a stiff adhesive will result in a larger changein output voltage and the gage factor will be higher Obtained results are consistent with the results ofthe strain transfer test presented in Section 2

For the sensing sheet bonded with Araldite sensors 2 and 7 showed high voltage output noise inthe change in output voltage values and therefore did not yield reasonable gage factor values For thesensing sheet bonded with 3M DP105 sensor 2 had high noise and did not yield reasonable gagefactor values Because change in output voltage values for sensor 2 were unreasonable in both theAraldite and 3M DP105 tests it is likely that there is a design or manufacturing flaw in the sensingsheet itself or a soldering problem on the rigid printed circuit board (see Figure 15) In the case ofsensor 7 in the Araldite-bonded sensor sheet the high noise can likely be attributed to a manufacturingflaw on the sensor sheet itself

Figure 16 shows that there is a range in apparent gage factor values (dispersion) from sensorto sensor One reason the gage factors might vary from sensor to sensor has to do with a possiblevoltage drop from the sensor to the rigid printed circuit board and reading unit Because the wiresin the sensing sheet have high resistance sensors that are located farthest from the reading unit(sensors 0 and 4 for example) may show a lower voltage output change than sensors that are closerto the reading unit (sensors 3 and 8) Furthermore a lower voltage output change will result in alower gage factor Figure 16 shows generally lower gage factors for sensors 0 and 4 validating thishypothesis Sensors 1 and 5 show higher gage factors and exhibit similar behavior Sensor 7 doesnot follow the trend described which may be attributed to a manufacturing flaw Another reasonfor varying gage factors between sensors is that the adhesive layerrsquos thickness might vary across thesensor sheet A thinner adhesive layer will permit more strain to be transferred to the sensor and thegage factor will be higher Conversely a thicker adhesive layer will transfer less strain to the sensorand will result in a lower computed gage factor

Finally Figure 17 and Table 1 show that the Araldite-bonded sensing sheet yields greater gagefactor uncertainties than the 3M DP105-bonded sensing sheet The Araldite adhesive is much moreviscous than the 3M DP105 adhesive and did not spread as easily when bonding the sensor sheet to thealuminum beam resulting in minor air pockets under the sensor sheet With the 3M DP105 adhesiveit was much easier to spread an even layer on the aluminum beam and no air pockets were visiblelikely increasing apparent gage factor consistency both between and within sensor sheets

4 Conclusions

The aim of this paper was to determine a suitable adhesive for installation of a novel sensing sheetonto steel structural members subjected to cracking Previous research has shown that for sensingsheets installed using a stiff adhesive (eg Araldite) the cracking in steel would induce immediatefailure of individual sensors of the sheet To address this challenge three soft 3M glues were consideredDP100 DP105 and DP125

The 3M DP125 adhesive was eliminated because the set time was too long to be useful inpractical applications Strain transfer tests were performed using strain gages to assess the quality

Sensors 2018 18 1907 15 of 16

of strain transfer of the two other glues It was shown that while the Araldite adhesive transfersa large percentage of strain to the sensors bonding with both the 3M DP100 and DP105 providessufficient sensitivity to detect small strain changes even in the non-damage range No conclusiveresults were found to justify the preference between the DP100 or DP105 from the non-damage studyso crack opening damage tests were performed 48-hour and 72-hour cure tests showed that the DP105adhesive was able to detect strain changes for larger crack opening sizes compared to the DP100 so itwas deemed more suitable for damage detection (cracking) in mechanically non-degrading materials(eg metals)

A sensing sheet prototype manufactured at Princeton University was then tested using the 3MDP105 adhesive For each of the working individual sensors apparent gage factors were determinedby loading and unloading a thin aluminum beam with sand bottles using theoretical strain valuesunder Euler-Bernoulli beam theory Uncertainties in the apparent gage factor values were found forboth Araldite and 3M DP105-bonded sensor sheets As expected the apparent gage factors for theAraldite adhesive were higher than for the 3M DP105 adhesive We know that an approximate gagefactor for bonding a sensor sheet with 3M DP105 adhesive would allow one to use the sensor sheet toidentify real strain values on structures

The consistency between the strain transfer tests performed on regular strain gages and the testsperformed on sensing sheets confirm that DP105 can be used for the installation of sensing sheets onsteel structural members for crack detection and characterization purposes

Author Contributions BG NV and JES created the concept of sensing sheet MG designed and carried outthe tests and data analysis he is the lead author of the paper CW and LEA developed and manufacturedsensing sheet prototype and associated hardware

Acknowledgments This research was in part supported by the Princeton Institute for the Science andTechnology of Materials (PRISM) and in part by the USDOT-RITA UTC Program grant no DTRT12-G- UTC16enabled through the Center for Advanced Infrastructure and Transportation (CAIT) at Rutgers UniversityThe authors would like to thank Y Hu L Huang N Lin W Rieutort-Louis J Sanz-Robinson T Liu S Wagnerand J Vocaturo from Princeton University for their precious help Kaitlyn Kliever helped improve the contents ofthis paper

Conflicts of Interest The authors declare no conflict of interest

References

1 Yao Y Glišic B Reliable damage detection and localization using direct strain sensing In Proceedings ofthe Sixth International IAMBAS Conference on Bridge Maintenance Safety and Management Stresa Italy8ndash12 July 2012 pp 714ndash721

2 Yao Y Tung S-TE Glišic B Crack detection and characterization techniquesmdashAn overview Struct ControlHealth Monit 2014 21 1387ndash1413 [CrossRef]

3 Tung S-T Yao Y Glišic B Sensing Sheet The Sensitivity of Thin-Film Full- Bridge Strain Sensors for CrackDetection and Characterization Meas Sci Technol 2014 25 14 [CrossRef]

4 Glišic B Chen J Hubbell D Streicker Bridge A comparison between Bragg-gratings long-gaugestrain and temperature sensors and Brillouin scattering-based distributed strain and temperature sensorsIn Proceedings of the SPIE Smart StructuresNDE and Materials + Nondestructive Evaluation and HealthMonitoring San Diego CA USA 6ndash10 March 2011

5 Loh KJ Hou TC Lynch JP Kotov NA Carbon Nanotube Sensing Skins for Spatial Strain and ImpactDamage Identification J Nondestruct Eval 2009 28 9ndash25 [CrossRef]

6 Pyo S Loh KJ Hou TC Jarva E Lynch JP A wireless impedance analyzer for automated tomographicmapping of a nanoengineered sensing skin Smart Struct Syst 2011 8 139ndash155 [CrossRef]

7 Pour-Ghaz M Weiss J Application of Frequency Selective Circuits for Crack Detection in ConcreteElements J ASTM Int 2011 8 1ndash11

8 Schumacher T Thostenson ET Development of structural carbon nanotube-based sensing composites forconcrete structures J Intell Mater Syst Struct 2014 25 1331ndash1339 [CrossRef]

Sensors 2018 18 1907 6 of 16

Figure 5 Strain length vs sample separation on non-degrading materials

22 Qualitative Tests

Qualitative observations were performed for three high-strength flexible adhesive alternatives3M DP100 DP105 and DP125 Initially these three adhesives were tested on paper samples in orderto gain an understanding of viscosity flexibility and set time It was found that the three adhesiveslisted were very flexible and could be bent and twisted (along with the substrate paper sheet) byhand without much effort A sample of a flexible adhesive applied to paper and bent can be seenin Figure 6 below

Figure 6 Sample of a flexible adhesive applied to paper

The 3M DP100 had a set time of about five minutes the DP105 had a set time of about 20 minutesand the DP125 had a set time which exceeded half an hour An ideal set time is neither too short(to give adequate time to place the sensor sheet) nor too long (to avoid sensor displacement caused byaccidental contact shortly after placement) A set time exceeding a half hour is undesirable because itincreases the risk of accidentally moving or removing the sensor during the installation phase Becausethe DP125 adhesive set time exceeded half an hour it was eliminated and the DP100 and DP105adhesives were further tested

Sensors 2018 18 1907 7 of 16

23 Quantitative Tests Loading and Unloading

Strain tests were performed in order to understand how much strain is transferred to the sensorfrom a monitored material given that the adhesives are soft and some loss of strain transfer is expectedThe test involved bonding three sensors to the top of a thin aluminum simply supported beam andloading and unloading the beam with known point forces Due to limitations of available testingset-up bottles filled with sand were used to load the beam The beam was simply supported at bothends as shown in Figure 7 It had span of 1710 mm thickness of 913 mm and width of 254 mmThe beam was made of aluminum and had a Youngrsquos modulus of 68 GPa Strain gages were placed at855 mm from the left support (in the middle of the beam) The sand bottles B1 B2 and B3 had masses5942 g (58 kN) 5981 g (59 kN) and 5977 g (59 kN) respectively For each load step the strain valuewas recorded with an Omega DP41-S reading unit (Omega Engineering Stamford CT USA) The totalloading and unloading time was less than 2 minutes The three tested adhesives were a stiff Araldite2012 adhesive (used by Yao and Glišic in previous studies [16]) 3M DP100 and 3M DP105 The stiffadhesive was used for reference purposes as it is assumed to have high quantity of strain transferdue to its high stiffness The aluminum beam and three bonded sensors used in the test are shown inFigure 7 below

Figure 7 Strain transfer test set-up (a) view of beam without sensors (b) support detail (c) view ofsensors installed using Araldite 2012 3M DP 100 and 3M DP 105 (d) sensor details

The results of the strain transfer tests are shown in Figure 8 An analytic strain model wasdeveloped for the aluminum beam The model was based on the linear theory of beams and ispresented in Equation (5)

εs =M times ys

EI(5)

where M = FL24mdashbending moment at location of the sensors (F is force applied using bottles placedin the middle of the beam L = 1710 mm is length of the beam)

E = 68 GPamdashYoungrsquos modulus of aluminumI = bh312mdashmoment of inertia of the beamrsquos cross-section (b = 254 mm h = 913 mm)

and ys = h2 + 02 mmmdashdistance of sensors from the horizontal principal axis of the cross-section(where 02 mm accounts for estimated thickness of the glue plus half the thickness of the strain gage)

Sensors 2018 18 1907 8 of 16

Figure 8 shows the instantaneous bending strain for seven load steps The first and last load stepscorrespond with no beam loading the second and sixth load steps correspond with the beam loadedwith sand bottle B1 the third and fifth load steps correspond with the beam loaded with sand bottlesB1 and B2 and the fourth load step corresponds with the beam loaded with all three sand bottlesAs expected the stiff Araldite adhesive consistently transfers the greatest fraction of strain from thestructure to the sensor and none of the adhesives transferred all strain to the sensor The strain valuesmeasured by sensors glued with Araldite were about 25 times greater than those measured by theother two glues Figure 8 shows that while DP100 transfers slightly more strain than DP105 in generalthere is no significant strain transfer difference between the 3M DP100 and 3M DP105 adhesivesBoth adhesives ldquoloserdquo more than 50 percent of strain however this is acceptable as the main purposeof the sensing sheet is damage detection and not strain monitoring Moreover linear response ofboth adhesives indicates that lack in strain transfer can be accounted for through calibration and gagefactor adjustments This is effectively done in Section 3 of this paper where strain transfer was testedon sensing sheet prototypes

Figure 8 Strain transfer results for various adhesives

24 Damage Tests Crack Opening vs Strain Test

The aim of the damage test was to assess the behavior of adhesives at crack locations andto see how it affects the survival of the sensors For these tests a sensor was bonded to ldquoclosedrdquoaluminum plates and the adhesive was given time to cure Once the adhesive reached suitablestrength an artificial crack was generated by pulling the aluminum plates apart thereby creating a gapbetween them The test setup is shown in Figure 9 Each simulated crack opening increment involvedpulling the aluminum plates apart by 0001 inches taking the strain reading and then waiting fiveminutes to record the ldquorelaxedrdquo strain reading Relaxation is expected due to creep in sensor polyimidesubstrate as well as due to creep in the adhesive exposed to high stress levels This process wasrepeated until failure (ie until the sensor could no longer read the strain due to rupture or the strainvalues began to decrease due to debonding or the adhesive breaking) The tests were performed withVishay 3800 strain reading units and the gage factor was calibrated to the manufacturerrsquos specificationsAn example of post-test damage is shown below in Figure 9

Sensors 2018 18 1907 9 of 16

Figure 9 Crack test setup

Based on the manufacturerrsquos specifications both adhesives achieve 80 of their full strengths ifcured for 20 minutes at 23 C and they achieve full strength if cured for 48 hours at 23 C Given thatthe temperature of the lab in which the tests were made was set to 17 C and could not be brought upto 23 C it was decided to perform one test before the full strength was achieved and one after the fullstrength was achieved For practical reasons the former was performed 24 hours after gluing and thelatter 72 hours after gluing (one day longer than specified time of 48 hours in order to account forlower curing temperature)

The following plots show the relationship between crack opening size and strain reading forthe 3M DP100 adhesive Both of these tests were performed with strain gages with a gage factor of201 In the first test (Figure 10) the adhesive was allowed to set for 24 hours and in the second testfor 72 hours (Figure 11) Time-dependent tests were performed in order to evaluate the adhesivesrsquoperformance with time and to establish a minimum cure time for future lab tests

Figure 10 DP100 crack opening vs strain after 24 hours

Sensors 2018 18 1907 10 of 16

Figure 11 DP100 crack opening vs strain after 72 hours

The time-dependent tests for the 3M DP100 adhesive showed very interesting results The testsdemonstrated that the bond strength does show time-dependence The 24-hour DP100 test showedreadings up to a crack opening of about 015 mm whereas the 72-hour test showed readings up toa 026 mm crack The result is important because it verifies that the adhesive performs better whengiven a few days to cure

Similar tests were performed for the 3M DP105 adhesive The results are shown inFigures 12 and 13 below

Figure 12 DP105 crack opening vs strain after 24 hours

Sensors 2018 18 1907 11 of 16

Figure 13 DP105 crack opening vs strain after 72 hours

The DP105 adhesive continued to read strain for crack openings up to 045 mm compared to about026 mm for the DP100 adhesive Taking into account that all other properties of the two adhesivesare similar the DP105 adhesive is considered the more suitable adhesive for the installation of sensingsheets that are supposed to detect and survive structural damage

3 Application to a Sensing Sheet Prototype

31 Objective

Once the appropriate adhesive for the sensing sheet was selected its application was testedon a sensing sheet prototype Given that manufacturing of a sensing sheet at the prototype stageis expensive it was decided to perform only non-damage strain transfer tests so that the sensing sheetwould not be damaged and could be used for future research related to hardware

The first step in implementing the sensing sheet is to determine appropriate gage factors of theindividual sensors and observe how these gage factors change depending on the adhesive used tobond the sheet The same testing setup as described in Section 23 was used The sensing sheet wasinstalled in the third quarter of the beam ie 1282 mm from the left support The load was applied1069 mm from the left support

The prototype of the sensing sheet is composed of eight individual sensors each composed offour resistors forming a full bridge Each individual sensor was independently tested by bonding thesensing sheet to an aluminum beam then loading and unloading the beam with sand bottles similarto the strain-transfer test described in the previous section In total two sensing sheets were testedthe first glued with the stiff Araldite adhesive for reference purposes and the second glued with thesoft DP105 glue Each loading step lasted 15 s and values were sampled at 1000 Hz 10 seconds intoeach loading step The beam was loaded for seven loading steps no load B1 B1 and B2 B1 and B2and B3 B1 and B2 B1 and no load giving a total testing time of 105 seconds At each loading step thechange in output voltage was recorded Using the change in output voltage the input voltage and ananalytical strain model the gage factor was determined for each sensor under each loading conditionFor each sensor identical tests were performed three times yielding three gage factors Preliminarytests included air and beam surface temperature measurements but no significant temperature changeswere found during the test duration Figures 14 and 15 show the test setup

Sensors 2018 18 1907 12 of 16

Figure 14 Sensing sheet and individual sensor numbering convention

Figure 15 Gage factor test setup The sensing sheet is bonded to the top surface of the aluminum beamand is connected to the rigid printed circuit board The temperature reading device is shown in thebottom left corner

32 Results

For each loading step for each individual sensor apparent gage factors were determined underidentical loading conditions and they are shown in Figure 16 Note that these gage factors are apparentand not true as they combine true gage factor with strain transfer quality The figure shows thatconsistent results are obtained for all individual sensors within the same sheet In addition the gagefactors for sensors attached using the DP105 adhesive are about 25 times less than those for the sensorsattached using Araldite (see Table 1) This result is consistent with strain transfer test results describedearlier in Section 2

As expected smaller loads yielded greater gage factor errors than large loads due to lower signalto noise ratios This created dispersion in the results Figure 17 plots the gage factor uncertaintymeasured as standard deviation for each load step Indices 1 and 5 represent the beam loaded withbottle B1 2 and 4 represent loading with bottles B1 and B2 and 3 represents loading with all three

Sensors 2018 18 1907 13 of 16

bottles Loading step 3 has the smallest gage factor uncertainty with a standard deviation of about006 for 3M DP105 and 013 for Araldite Loading step 5 has the greatest gage factor uncertaintywith a standard deviation of about 033 for 3M DP105 and 054 for Araldite As Figure 17 showsthe Araldite and 3M DP105 have similar gage factor uncertainty trends except the Aralditersquos gagefactor uncertainties are consistently greater than those for 3M DP105 The reason for this could be theimperfections and connection issues in the flexible to rigid interface where a zero-insertion-force (ZIF)connector is used Unwanted impedance in this connection can vary by insertion and introduces asmall amount of noise into the differential voltage which is used to calculate strain This also showsthe need for an integrated system implementation on a flexible substrate and the need to reduce thenumber of interfaces to rigid boards as mentioned in previous works [18ndash20]

Figure 16 Apparent gage factors determined on individual sensors of sensing sheet prototypes gluedwith Araldite and 3M DP105 adhesives

Figure 17 Apparent gage factor uncertainties for various load steps

Sensors 2018 18 1907 14 of 16

Table 1 Apparent gage factor summary for Araldite and 3M DP105 adhesives (microlowast = meanσlowast = standard deviation)

Araldite GF 3M DP105 GF

microlowast 135 054σlowast 0240 0128

33 Discussion

The apparent gage factor for Araldite is greater than the apparent gage factor for 3M DP105adhesive because the Araldite is less flexible and transfers more strain from the structure to the sensorStrain is proportional to change in output voltage and inversely proportional to gage factor so if ananalytical strain model is used to determine gage factor a stiff adhesive will result in a larger changein output voltage and the gage factor will be higher Obtained results are consistent with the results ofthe strain transfer test presented in Section 2

For the sensing sheet bonded with Araldite sensors 2 and 7 showed high voltage output noise inthe change in output voltage values and therefore did not yield reasonable gage factor values For thesensing sheet bonded with 3M DP105 sensor 2 had high noise and did not yield reasonable gagefactor values Because change in output voltage values for sensor 2 were unreasonable in both theAraldite and 3M DP105 tests it is likely that there is a design or manufacturing flaw in the sensingsheet itself or a soldering problem on the rigid printed circuit board (see Figure 15) In the case ofsensor 7 in the Araldite-bonded sensor sheet the high noise can likely be attributed to a manufacturingflaw on the sensor sheet itself

Figure 16 shows that there is a range in apparent gage factor values (dispersion) from sensorto sensor One reason the gage factors might vary from sensor to sensor has to do with a possiblevoltage drop from the sensor to the rigid printed circuit board and reading unit Because the wiresin the sensing sheet have high resistance sensors that are located farthest from the reading unit(sensors 0 and 4 for example) may show a lower voltage output change than sensors that are closerto the reading unit (sensors 3 and 8) Furthermore a lower voltage output change will result in alower gage factor Figure 16 shows generally lower gage factors for sensors 0 and 4 validating thishypothesis Sensors 1 and 5 show higher gage factors and exhibit similar behavior Sensor 7 doesnot follow the trend described which may be attributed to a manufacturing flaw Another reasonfor varying gage factors between sensors is that the adhesive layerrsquos thickness might vary across thesensor sheet A thinner adhesive layer will permit more strain to be transferred to the sensor and thegage factor will be higher Conversely a thicker adhesive layer will transfer less strain to the sensorand will result in a lower computed gage factor

Finally Figure 17 and Table 1 show that the Araldite-bonded sensing sheet yields greater gagefactor uncertainties than the 3M DP105-bonded sensing sheet The Araldite adhesive is much moreviscous than the 3M DP105 adhesive and did not spread as easily when bonding the sensor sheet to thealuminum beam resulting in minor air pockets under the sensor sheet With the 3M DP105 adhesiveit was much easier to spread an even layer on the aluminum beam and no air pockets were visiblelikely increasing apparent gage factor consistency both between and within sensor sheets

4 Conclusions

The aim of this paper was to determine a suitable adhesive for installation of a novel sensing sheetonto steel structural members subjected to cracking Previous research has shown that for sensingsheets installed using a stiff adhesive (eg Araldite) the cracking in steel would induce immediatefailure of individual sensors of the sheet To address this challenge three soft 3M glues were consideredDP100 DP105 and DP125

The 3M DP125 adhesive was eliminated because the set time was too long to be useful inpractical applications Strain transfer tests were performed using strain gages to assess the quality

Sensors 2018 18 1907 15 of 16

of strain transfer of the two other glues It was shown that while the Araldite adhesive transfersa large percentage of strain to the sensors bonding with both the 3M DP100 and DP105 providessufficient sensitivity to detect small strain changes even in the non-damage range No conclusiveresults were found to justify the preference between the DP100 or DP105 from the non-damage studyso crack opening damage tests were performed 48-hour and 72-hour cure tests showed that the DP105adhesive was able to detect strain changes for larger crack opening sizes compared to the DP100 so itwas deemed more suitable for damage detection (cracking) in mechanically non-degrading materials(eg metals)

A sensing sheet prototype manufactured at Princeton University was then tested using the 3MDP105 adhesive For each of the working individual sensors apparent gage factors were determinedby loading and unloading a thin aluminum beam with sand bottles using theoretical strain valuesunder Euler-Bernoulli beam theory Uncertainties in the apparent gage factor values were found forboth Araldite and 3M DP105-bonded sensor sheets As expected the apparent gage factors for theAraldite adhesive were higher than for the 3M DP105 adhesive We know that an approximate gagefactor for bonding a sensor sheet with 3M DP105 adhesive would allow one to use the sensor sheet toidentify real strain values on structures

The consistency between the strain transfer tests performed on regular strain gages and the testsperformed on sensing sheets confirm that DP105 can be used for the installation of sensing sheets onsteel structural members for crack detection and characterization purposes

Author Contributions BG NV and JES created the concept of sensing sheet MG designed and carried outthe tests and data analysis he is the lead author of the paper CW and LEA developed and manufacturedsensing sheet prototype and associated hardware

Acknowledgments This research was in part supported by the Princeton Institute for the Science andTechnology of Materials (PRISM) and in part by the USDOT-RITA UTC Program grant no DTRT12-G- UTC16enabled through the Center for Advanced Infrastructure and Transportation (CAIT) at Rutgers UniversityThe authors would like to thank Y Hu L Huang N Lin W Rieutort-Louis J Sanz-Robinson T Liu S Wagnerand J Vocaturo from Princeton University for their precious help Kaitlyn Kliever helped improve the contents ofthis paper

Conflicts of Interest The authors declare no conflict of interest

References

1 Yao Y Glišic B Reliable damage detection and localization using direct strain sensing In Proceedings ofthe Sixth International IAMBAS Conference on Bridge Maintenance Safety and Management Stresa Italy8ndash12 July 2012 pp 714ndash721

2 Yao Y Tung S-TE Glišic B Crack detection and characterization techniquesmdashAn overview Struct ControlHealth Monit 2014 21 1387ndash1413 [CrossRef]

3 Tung S-T Yao Y Glišic B Sensing Sheet The Sensitivity of Thin-Film Full- Bridge Strain Sensors for CrackDetection and Characterization Meas Sci Technol 2014 25 14 [CrossRef]

4 Glišic B Chen J Hubbell D Streicker Bridge A comparison between Bragg-gratings long-gaugestrain and temperature sensors and Brillouin scattering-based distributed strain and temperature sensorsIn Proceedings of the SPIE Smart StructuresNDE and Materials + Nondestructive Evaluation and HealthMonitoring San Diego CA USA 6ndash10 March 2011

5 Loh KJ Hou TC Lynch JP Kotov NA Carbon Nanotube Sensing Skins for Spatial Strain and ImpactDamage Identification J Nondestruct Eval 2009 28 9ndash25 [CrossRef]

6 Pyo S Loh KJ Hou TC Jarva E Lynch JP A wireless impedance analyzer for automated tomographicmapping of a nanoengineered sensing skin Smart Struct Syst 2011 8 139ndash155 [CrossRef]

7 Pour-Ghaz M Weiss J Application of Frequency Selective Circuits for Crack Detection in ConcreteElements J ASTM Int 2011 8 1ndash11

8 Schumacher T Thostenson ET Development of structural carbon nanotube-based sensing composites forconcrete structures J Intell Mater Syst Struct 2014 25 1331ndash1339 [CrossRef]

Sensors 2018 18 1907 7 of 16

23 Quantitative Tests Loading and Unloading

Strain tests were performed in order to understand how much strain is transferred to the sensorfrom a monitored material given that the adhesives are soft and some loss of strain transfer is expectedThe test involved bonding three sensors to the top of a thin aluminum simply supported beam andloading and unloading the beam with known point forces Due to limitations of available testingset-up bottles filled with sand were used to load the beam The beam was simply supported at bothends as shown in Figure 7 It had span of 1710 mm thickness of 913 mm and width of 254 mmThe beam was made of aluminum and had a Youngrsquos modulus of 68 GPa Strain gages were placed at855 mm from the left support (in the middle of the beam) The sand bottles B1 B2 and B3 had masses5942 g (58 kN) 5981 g (59 kN) and 5977 g (59 kN) respectively For each load step the strain valuewas recorded with an Omega DP41-S reading unit (Omega Engineering Stamford CT USA) The totalloading and unloading time was less than 2 minutes The three tested adhesives were a stiff Araldite2012 adhesive (used by Yao and Glišic in previous studies [16]) 3M DP100 and 3M DP105 The stiffadhesive was used for reference purposes as it is assumed to have high quantity of strain transferdue to its high stiffness The aluminum beam and three bonded sensors used in the test are shown inFigure 7 below

Figure 7 Strain transfer test set-up (a) view of beam without sensors (b) support detail (c) view ofsensors installed using Araldite 2012 3M DP 100 and 3M DP 105 (d) sensor details

The results of the strain transfer tests are shown in Figure 8 An analytic strain model wasdeveloped for the aluminum beam The model was based on the linear theory of beams and ispresented in Equation (5)

εs =M times ys

EI(5)

where M = FL24mdashbending moment at location of the sensors (F is force applied using bottles placedin the middle of the beam L = 1710 mm is length of the beam)

E = 68 GPamdashYoungrsquos modulus of aluminumI = bh312mdashmoment of inertia of the beamrsquos cross-section (b = 254 mm h = 913 mm)

and ys = h2 + 02 mmmdashdistance of sensors from the horizontal principal axis of the cross-section(where 02 mm accounts for estimated thickness of the glue plus half the thickness of the strain gage)

Sensors 2018 18 1907 8 of 16

Figure 8 shows the instantaneous bending strain for seven load steps The first and last load stepscorrespond with no beam loading the second and sixth load steps correspond with the beam loadedwith sand bottle B1 the third and fifth load steps correspond with the beam loaded with sand bottlesB1 and B2 and the fourth load step corresponds with the beam loaded with all three sand bottlesAs expected the stiff Araldite adhesive consistently transfers the greatest fraction of strain from thestructure to the sensor and none of the adhesives transferred all strain to the sensor The strain valuesmeasured by sensors glued with Araldite were about 25 times greater than those measured by theother two glues Figure 8 shows that while DP100 transfers slightly more strain than DP105 in generalthere is no significant strain transfer difference between the 3M DP100 and 3M DP105 adhesivesBoth adhesives ldquoloserdquo more than 50 percent of strain however this is acceptable as the main purposeof the sensing sheet is damage detection and not strain monitoring Moreover linear response ofboth adhesives indicates that lack in strain transfer can be accounted for through calibration and gagefactor adjustments This is effectively done in Section 3 of this paper where strain transfer was testedon sensing sheet prototypes

Figure 8 Strain transfer results for various adhesives

24 Damage Tests Crack Opening vs Strain Test

The aim of the damage test was to assess the behavior of adhesives at crack locations andto see how it affects the survival of the sensors For these tests a sensor was bonded to ldquoclosedrdquoaluminum plates and the adhesive was given time to cure Once the adhesive reached suitablestrength an artificial crack was generated by pulling the aluminum plates apart thereby creating a gapbetween them The test setup is shown in Figure 9 Each simulated crack opening increment involvedpulling the aluminum plates apart by 0001 inches taking the strain reading and then waiting fiveminutes to record the ldquorelaxedrdquo strain reading Relaxation is expected due to creep in sensor polyimidesubstrate as well as due to creep in the adhesive exposed to high stress levels This process wasrepeated until failure (ie until the sensor could no longer read the strain due to rupture or the strainvalues began to decrease due to debonding or the adhesive breaking) The tests were performed withVishay 3800 strain reading units and the gage factor was calibrated to the manufacturerrsquos specificationsAn example of post-test damage is shown below in Figure 9

Sensors 2018 18 1907 9 of 16

Figure 9 Crack test setup

Based on the manufacturerrsquos specifications both adhesives achieve 80 of their full strengths ifcured for 20 minutes at 23 C and they achieve full strength if cured for 48 hours at 23 C Given thatthe temperature of the lab in which the tests were made was set to 17 C and could not be brought upto 23 C it was decided to perform one test before the full strength was achieved and one after the fullstrength was achieved For practical reasons the former was performed 24 hours after gluing and thelatter 72 hours after gluing (one day longer than specified time of 48 hours in order to account forlower curing temperature)

The following plots show the relationship between crack opening size and strain reading forthe 3M DP100 adhesive Both of these tests were performed with strain gages with a gage factor of201 In the first test (Figure 10) the adhesive was allowed to set for 24 hours and in the second testfor 72 hours (Figure 11) Time-dependent tests were performed in order to evaluate the adhesivesrsquoperformance with time and to establish a minimum cure time for future lab tests

Figure 10 DP100 crack opening vs strain after 24 hours

Sensors 2018 18 1907 10 of 16

Figure 11 DP100 crack opening vs strain after 72 hours

The time-dependent tests for the 3M DP100 adhesive showed very interesting results The testsdemonstrated that the bond strength does show time-dependence The 24-hour DP100 test showedreadings up to a crack opening of about 015 mm whereas the 72-hour test showed readings up toa 026 mm crack The result is important because it verifies that the adhesive performs better whengiven a few days to cure

Similar tests were performed for the 3M DP105 adhesive The results are shown inFigures 12 and 13 below

Figure 12 DP105 crack opening vs strain after 24 hours

Sensors 2018 18 1907 11 of 16

Figure 13 DP105 crack opening vs strain after 72 hours

The DP105 adhesive continued to read strain for crack openings up to 045 mm compared to about026 mm for the DP100 adhesive Taking into account that all other properties of the two adhesivesare similar the DP105 adhesive is considered the more suitable adhesive for the installation of sensingsheets that are supposed to detect and survive structural damage

3 Application to a Sensing Sheet Prototype

31 Objective

Once the appropriate adhesive for the sensing sheet was selected its application was testedon a sensing sheet prototype Given that manufacturing of a sensing sheet at the prototype stageis expensive it was decided to perform only non-damage strain transfer tests so that the sensing sheetwould not be damaged and could be used for future research related to hardware

The first step in implementing the sensing sheet is to determine appropriate gage factors of theindividual sensors and observe how these gage factors change depending on the adhesive used tobond the sheet The same testing setup as described in Section 23 was used The sensing sheet wasinstalled in the third quarter of the beam ie 1282 mm from the left support The load was applied1069 mm from the left support

The prototype of the sensing sheet is composed of eight individual sensors each composed offour resistors forming a full bridge Each individual sensor was independently tested by bonding thesensing sheet to an aluminum beam then loading and unloading the beam with sand bottles similarto the strain-transfer test described in the previous section In total two sensing sheets were testedthe first glued with the stiff Araldite adhesive for reference purposes and the second glued with thesoft DP105 glue Each loading step lasted 15 s and values were sampled at 1000 Hz 10 seconds intoeach loading step The beam was loaded for seven loading steps no load B1 B1 and B2 B1 and B2and B3 B1 and B2 B1 and no load giving a total testing time of 105 seconds At each loading step thechange in output voltage was recorded Using the change in output voltage the input voltage and ananalytical strain model the gage factor was determined for each sensor under each loading conditionFor each sensor identical tests were performed three times yielding three gage factors Preliminarytests included air and beam surface temperature measurements but no significant temperature changeswere found during the test duration Figures 14 and 15 show the test setup

Sensors 2018 18 1907 12 of 16

Figure 14 Sensing sheet and individual sensor numbering convention

Figure 15 Gage factor test setup The sensing sheet is bonded to the top surface of the aluminum beamand is connected to the rigid printed circuit board The temperature reading device is shown in thebottom left corner

32 Results

For each loading step for each individual sensor apparent gage factors were determined underidentical loading conditions and they are shown in Figure 16 Note that these gage factors are apparentand not true as they combine true gage factor with strain transfer quality The figure shows thatconsistent results are obtained for all individual sensors within the same sheet In addition the gagefactors for sensors attached using the DP105 adhesive are about 25 times less than those for the sensorsattached using Araldite (see Table 1) This result is consistent with strain transfer test results describedearlier in Section 2

As expected smaller loads yielded greater gage factor errors than large loads due to lower signalto noise ratios This created dispersion in the results Figure 17 plots the gage factor uncertaintymeasured as standard deviation for each load step Indices 1 and 5 represent the beam loaded withbottle B1 2 and 4 represent loading with bottles B1 and B2 and 3 represents loading with all three

Sensors 2018 18 1907 13 of 16

bottles Loading step 3 has the smallest gage factor uncertainty with a standard deviation of about006 for 3M DP105 and 013 for Araldite Loading step 5 has the greatest gage factor uncertaintywith a standard deviation of about 033 for 3M DP105 and 054 for Araldite As Figure 17 showsthe Araldite and 3M DP105 have similar gage factor uncertainty trends except the Aralditersquos gagefactor uncertainties are consistently greater than those for 3M DP105 The reason for this could be theimperfections and connection issues in the flexible to rigid interface where a zero-insertion-force (ZIF)connector is used Unwanted impedance in this connection can vary by insertion and introduces asmall amount of noise into the differential voltage which is used to calculate strain This also showsthe need for an integrated system implementation on a flexible substrate and the need to reduce thenumber of interfaces to rigid boards as mentioned in previous works [18ndash20]

Figure 16 Apparent gage factors determined on individual sensors of sensing sheet prototypes gluedwith Araldite and 3M DP105 adhesives

Figure 17 Apparent gage factor uncertainties for various load steps

Sensors 2018 18 1907 14 of 16

Table 1 Apparent gage factor summary for Araldite and 3M DP105 adhesives (microlowast = meanσlowast = standard deviation)

Araldite GF 3M DP105 GF

microlowast 135 054σlowast 0240 0128

33 Discussion

The apparent gage factor for Araldite is greater than the apparent gage factor for 3M DP105adhesive because the Araldite is less flexible and transfers more strain from the structure to the sensorStrain is proportional to change in output voltage and inversely proportional to gage factor so if ananalytical strain model is used to determine gage factor a stiff adhesive will result in a larger changein output voltage and the gage factor will be higher Obtained results are consistent with the results ofthe strain transfer test presented in Section 2

For the sensing sheet bonded with Araldite sensors 2 and 7 showed high voltage output noise inthe change in output voltage values and therefore did not yield reasonable gage factor values For thesensing sheet bonded with 3M DP105 sensor 2 had high noise and did not yield reasonable gagefactor values Because change in output voltage values for sensor 2 were unreasonable in both theAraldite and 3M DP105 tests it is likely that there is a design or manufacturing flaw in the sensingsheet itself or a soldering problem on the rigid printed circuit board (see Figure 15) In the case ofsensor 7 in the Araldite-bonded sensor sheet the high noise can likely be attributed to a manufacturingflaw on the sensor sheet itself

Figure 16 shows that there is a range in apparent gage factor values (dispersion) from sensorto sensor One reason the gage factors might vary from sensor to sensor has to do with a possiblevoltage drop from the sensor to the rigid printed circuit board and reading unit Because the wiresin the sensing sheet have high resistance sensors that are located farthest from the reading unit(sensors 0 and 4 for example) may show a lower voltage output change than sensors that are closerto the reading unit (sensors 3 and 8) Furthermore a lower voltage output change will result in alower gage factor Figure 16 shows generally lower gage factors for sensors 0 and 4 validating thishypothesis Sensors 1 and 5 show higher gage factors and exhibit similar behavior Sensor 7 doesnot follow the trend described which may be attributed to a manufacturing flaw Another reasonfor varying gage factors between sensors is that the adhesive layerrsquos thickness might vary across thesensor sheet A thinner adhesive layer will permit more strain to be transferred to the sensor and thegage factor will be higher Conversely a thicker adhesive layer will transfer less strain to the sensorand will result in a lower computed gage factor

Finally Figure 17 and Table 1 show that the Araldite-bonded sensing sheet yields greater gagefactor uncertainties than the 3M DP105-bonded sensing sheet The Araldite adhesive is much moreviscous than the 3M DP105 adhesive and did not spread as easily when bonding the sensor sheet to thealuminum beam resulting in minor air pockets under the sensor sheet With the 3M DP105 adhesiveit was much easier to spread an even layer on the aluminum beam and no air pockets were visiblelikely increasing apparent gage factor consistency both between and within sensor sheets

4 Conclusions

The aim of this paper was to determine a suitable adhesive for installation of a novel sensing sheetonto steel structural members subjected to cracking Previous research has shown that for sensingsheets installed using a stiff adhesive (eg Araldite) the cracking in steel would induce immediatefailure of individual sensors of the sheet To address this challenge three soft 3M glues were consideredDP100 DP105 and DP125

The 3M DP125 adhesive was eliminated because the set time was too long to be useful inpractical applications Strain transfer tests were performed using strain gages to assess the quality

Sensors 2018 18 1907 15 of 16

of strain transfer of the two other glues It was shown that while the Araldite adhesive transfersa large percentage of strain to the sensors bonding with both the 3M DP100 and DP105 providessufficient sensitivity to detect small strain changes even in the non-damage range No conclusiveresults were found to justify the preference between the DP100 or DP105 from the non-damage studyso crack opening damage tests were performed 48-hour and 72-hour cure tests showed that the DP105adhesive was able to detect strain changes for larger crack opening sizes compared to the DP100 so itwas deemed more suitable for damage detection (cracking) in mechanically non-degrading materials(eg metals)

A sensing sheet prototype manufactured at Princeton University was then tested using the 3MDP105 adhesive For each of the working individual sensors apparent gage factors were determinedby loading and unloading a thin aluminum beam with sand bottles using theoretical strain valuesunder Euler-Bernoulli beam theory Uncertainties in the apparent gage factor values were found forboth Araldite and 3M DP105-bonded sensor sheets As expected the apparent gage factors for theAraldite adhesive were higher than for the 3M DP105 adhesive We know that an approximate gagefactor for bonding a sensor sheet with 3M DP105 adhesive would allow one to use the sensor sheet toidentify real strain values on structures

The consistency between the strain transfer tests performed on regular strain gages and the testsperformed on sensing sheets confirm that DP105 can be used for the installation of sensing sheets onsteel structural members for crack detection and characterization purposes

Author Contributions BG NV and JES created the concept of sensing sheet MG designed and carried outthe tests and data analysis he is the lead author of the paper CW and LEA developed and manufacturedsensing sheet prototype and associated hardware

Acknowledgments This research was in part supported by the Princeton Institute for the Science andTechnology of Materials (PRISM) and in part by the USDOT-RITA UTC Program grant no DTRT12-G- UTC16enabled through the Center for Advanced Infrastructure and Transportation (CAIT) at Rutgers UniversityThe authors would like to thank Y Hu L Huang N Lin W Rieutort-Louis J Sanz-Robinson T Liu S Wagnerand J Vocaturo from Princeton University for their precious help Kaitlyn Kliever helped improve the contents ofthis paper

Conflicts of Interest The authors declare no conflict of interest

References

1 Yao Y Glišic B Reliable damage detection and localization using direct strain sensing In Proceedings ofthe Sixth International IAMBAS Conference on Bridge Maintenance Safety and Management Stresa Italy8ndash12 July 2012 pp 714ndash721

2 Yao Y Tung S-TE Glišic B Crack detection and characterization techniquesmdashAn overview Struct ControlHealth Monit 2014 21 1387ndash1413 [CrossRef]

3 Tung S-T Yao Y Glišic B Sensing Sheet The Sensitivity of Thin-Film Full- Bridge Strain Sensors for CrackDetection and Characterization Meas Sci Technol 2014 25 14 [CrossRef]

4 Glišic B Chen J Hubbell D Streicker Bridge A comparison between Bragg-gratings long-gaugestrain and temperature sensors and Brillouin scattering-based distributed strain and temperature sensorsIn Proceedings of the SPIE Smart StructuresNDE and Materials + Nondestructive Evaluation and HealthMonitoring San Diego CA USA 6ndash10 March 2011

5 Loh KJ Hou TC Lynch JP Kotov NA Carbon Nanotube Sensing Skins for Spatial Strain and ImpactDamage Identification J Nondestruct Eval 2009 28 9ndash25 [CrossRef]

6 Pyo S Loh KJ Hou TC Jarva E Lynch JP A wireless impedance analyzer for automated tomographicmapping of a nanoengineered sensing skin Smart Struct Syst 2011 8 139ndash155 [CrossRef]

7 Pour-Ghaz M Weiss J Application of Frequency Selective Circuits for Crack Detection in ConcreteElements J ASTM Int 2011 8 1ndash11

8 Schumacher T Thostenson ET Development of structural carbon nanotube-based sensing composites forconcrete structures J Intell Mater Syst Struct 2014 25 1331ndash1339 [CrossRef]

Sensors 2018 18 1907 8 of 16

Figure 8 shows the instantaneous bending strain for seven load steps The first and last load stepscorrespond with no beam loading the second and sixth load steps correspond with the beam loadedwith sand bottle B1 the third and fifth load steps correspond with the beam loaded with sand bottlesB1 and B2 and the fourth load step corresponds with the beam loaded with all three sand bottlesAs expected the stiff Araldite adhesive consistently transfers the greatest fraction of strain from thestructure to the sensor and none of the adhesives transferred all strain to the sensor The strain valuesmeasured by sensors glued with Araldite were about 25 times greater than those measured by theother two glues Figure 8 shows that while DP100 transfers slightly more strain than DP105 in generalthere is no significant strain transfer difference between the 3M DP100 and 3M DP105 adhesivesBoth adhesives ldquoloserdquo more than 50 percent of strain however this is acceptable as the main purposeof the sensing sheet is damage detection and not strain monitoring Moreover linear response ofboth adhesives indicates that lack in strain transfer can be accounted for through calibration and gagefactor adjustments This is effectively done in Section 3 of this paper where strain transfer was testedon sensing sheet prototypes

Figure 8 Strain transfer results for various adhesives

24 Damage Tests Crack Opening vs Strain Test

The aim of the damage test was to assess the behavior of adhesives at crack locations andto see how it affects the survival of the sensors For these tests a sensor was bonded to ldquoclosedrdquoaluminum plates and the adhesive was given time to cure Once the adhesive reached suitablestrength an artificial crack was generated by pulling the aluminum plates apart thereby creating a gapbetween them The test setup is shown in Figure 9 Each simulated crack opening increment involvedpulling the aluminum plates apart by 0001 inches taking the strain reading and then waiting fiveminutes to record the ldquorelaxedrdquo strain reading Relaxation is expected due to creep in sensor polyimidesubstrate as well as due to creep in the adhesive exposed to high stress levels This process wasrepeated until failure (ie until the sensor could no longer read the strain due to rupture or the strainvalues began to decrease due to debonding or the adhesive breaking) The tests were performed withVishay 3800 strain reading units and the gage factor was calibrated to the manufacturerrsquos specificationsAn example of post-test damage is shown below in Figure 9

Sensors 2018 18 1907 9 of 16

Figure 9 Crack test setup

Based on the manufacturerrsquos specifications both adhesives achieve 80 of their full strengths ifcured for 20 minutes at 23 C and they achieve full strength if cured for 48 hours at 23 C Given thatthe temperature of the lab in which the tests were made was set to 17 C and could not be brought upto 23 C it was decided to perform one test before the full strength was achieved and one after the fullstrength was achieved For practical reasons the former was performed 24 hours after gluing and thelatter 72 hours after gluing (one day longer than specified time of 48 hours in order to account forlower curing temperature)

The following plots show the relationship between crack opening size and strain reading forthe 3M DP100 adhesive Both of these tests were performed with strain gages with a gage factor of201 In the first test (Figure 10) the adhesive was allowed to set for 24 hours and in the second testfor 72 hours (Figure 11) Time-dependent tests were performed in order to evaluate the adhesivesrsquoperformance with time and to establish a minimum cure time for future lab tests

Figure 10 DP100 crack opening vs strain after 24 hours

Sensors 2018 18 1907 10 of 16

Figure 11 DP100 crack opening vs strain after 72 hours

The time-dependent tests for the 3M DP100 adhesive showed very interesting results The testsdemonstrated that the bond strength does show time-dependence The 24-hour DP100 test showedreadings up to a crack opening of about 015 mm whereas the 72-hour test showed readings up toa 026 mm crack The result is important because it verifies that the adhesive performs better whengiven a few days to cure

Similar tests were performed for the 3M DP105 adhesive The results are shown inFigures 12 and 13 below

Figure 12 DP105 crack opening vs strain after 24 hours

Sensors 2018 18 1907 11 of 16

Figure 13 DP105 crack opening vs strain after 72 hours

The DP105 adhesive continued to read strain for crack openings up to 045 mm compared to about026 mm for the DP100 adhesive Taking into account that all other properties of the two adhesivesare similar the DP105 adhesive is considered the more suitable adhesive for the installation of sensingsheets that are supposed to detect and survive structural damage

3 Application to a Sensing Sheet Prototype

31 Objective

Once the appropriate adhesive for the sensing sheet was selected its application was testedon a sensing sheet prototype Given that manufacturing of a sensing sheet at the prototype stageis expensive it was decided to perform only non-damage strain transfer tests so that the sensing sheetwould not be damaged and could be used for future research related to hardware

The first step in implementing the sensing sheet is to determine appropriate gage factors of theindividual sensors and observe how these gage factors change depending on the adhesive used tobond the sheet The same testing setup as described in Section 23 was used The sensing sheet wasinstalled in the third quarter of the beam ie 1282 mm from the left support The load was applied1069 mm from the left support

The prototype of the sensing sheet is composed of eight individual sensors each composed offour resistors forming a full bridge Each individual sensor was independently tested by bonding thesensing sheet to an aluminum beam then loading and unloading the beam with sand bottles similarto the strain-transfer test described in the previous section In total two sensing sheets were testedthe first glued with the stiff Araldite adhesive for reference purposes and the second glued with thesoft DP105 glue Each loading step lasted 15 s and values were sampled at 1000 Hz 10 seconds intoeach loading step The beam was loaded for seven loading steps no load B1 B1 and B2 B1 and B2and B3 B1 and B2 B1 and no load giving a total testing time of 105 seconds At each loading step thechange in output voltage was recorded Using the change in output voltage the input voltage and ananalytical strain model the gage factor was determined for each sensor under each loading conditionFor each sensor identical tests were performed three times yielding three gage factors Preliminarytests included air and beam surface temperature measurements but no significant temperature changeswere found during the test duration Figures 14 and 15 show the test setup

Sensors 2018 18 1907 12 of 16

Figure 14 Sensing sheet and individual sensor numbering convention

Figure 15 Gage factor test setup The sensing sheet is bonded to the top surface of the aluminum beamand is connected to the rigid printed circuit board The temperature reading device is shown in thebottom left corner

32 Results

For each loading step for each individual sensor apparent gage factors were determined underidentical loading conditions and they are shown in Figure 16 Note that these gage factors are apparentand not true as they combine true gage factor with strain transfer quality The figure shows thatconsistent results are obtained for all individual sensors within the same sheet In addition the gagefactors for sensors attached using the DP105 adhesive are about 25 times less than those for the sensorsattached using Araldite (see Table 1) This result is consistent with strain transfer test results describedearlier in Section 2

As expected smaller loads yielded greater gage factor errors than large loads due to lower signalto noise ratios This created dispersion in the results Figure 17 plots the gage factor uncertaintymeasured as standard deviation for each load step Indices 1 and 5 represent the beam loaded withbottle B1 2 and 4 represent loading with bottles B1 and B2 and 3 represents loading with all three

Sensors 2018 18 1907 13 of 16

bottles Loading step 3 has the smallest gage factor uncertainty with a standard deviation of about006 for 3M DP105 and 013 for Araldite Loading step 5 has the greatest gage factor uncertaintywith a standard deviation of about 033 for 3M DP105 and 054 for Araldite As Figure 17 showsthe Araldite and 3M DP105 have similar gage factor uncertainty trends except the Aralditersquos gagefactor uncertainties are consistently greater than those for 3M DP105 The reason for this could be theimperfections and connection issues in the flexible to rigid interface where a zero-insertion-force (ZIF)connector is used Unwanted impedance in this connection can vary by insertion and introduces asmall amount of noise into the differential voltage which is used to calculate strain This also showsthe need for an integrated system implementation on a flexible substrate and the need to reduce thenumber of interfaces to rigid boards as mentioned in previous works [18ndash20]

Figure 16 Apparent gage factors determined on individual sensors of sensing sheet prototypes gluedwith Araldite and 3M DP105 adhesives

Figure 17 Apparent gage factor uncertainties for various load steps

Sensors 2018 18 1907 14 of 16

Table 1 Apparent gage factor summary for Araldite and 3M DP105 adhesives (microlowast = meanσlowast = standard deviation)

Araldite GF 3M DP105 GF

microlowast 135 054σlowast 0240 0128

33 Discussion

The apparent gage factor for Araldite is greater than the apparent gage factor for 3M DP105adhesive because the Araldite is less flexible and transfers more strain from the structure to the sensorStrain is proportional to change in output voltage and inversely proportional to gage factor so if ananalytical strain model is used to determine gage factor a stiff adhesive will result in a larger changein output voltage and the gage factor will be higher Obtained results are consistent with the results ofthe strain transfer test presented in Section 2

For the sensing sheet bonded with Araldite sensors 2 and 7 showed high voltage output noise inthe change in output voltage values and therefore did not yield reasonable gage factor values For thesensing sheet bonded with 3M DP105 sensor 2 had high noise and did not yield reasonable gagefactor values Because change in output voltage values for sensor 2 were unreasonable in both theAraldite and 3M DP105 tests it is likely that there is a design or manufacturing flaw in the sensingsheet itself or a soldering problem on the rigid printed circuit board (see Figure 15) In the case ofsensor 7 in the Araldite-bonded sensor sheet the high noise can likely be attributed to a manufacturingflaw on the sensor sheet itself

Figure 16 shows that there is a range in apparent gage factor values (dispersion) from sensorto sensor One reason the gage factors might vary from sensor to sensor has to do with a possiblevoltage drop from the sensor to the rigid printed circuit board and reading unit Because the wiresin the sensing sheet have high resistance sensors that are located farthest from the reading unit(sensors 0 and 4 for example) may show a lower voltage output change than sensors that are closerto the reading unit (sensors 3 and 8) Furthermore a lower voltage output change will result in alower gage factor Figure 16 shows generally lower gage factors for sensors 0 and 4 validating thishypothesis Sensors 1 and 5 show higher gage factors and exhibit similar behavior Sensor 7 doesnot follow the trend described which may be attributed to a manufacturing flaw Another reasonfor varying gage factors between sensors is that the adhesive layerrsquos thickness might vary across thesensor sheet A thinner adhesive layer will permit more strain to be transferred to the sensor and thegage factor will be higher Conversely a thicker adhesive layer will transfer less strain to the sensorand will result in a lower computed gage factor

Finally Figure 17 and Table 1 show that the Araldite-bonded sensing sheet yields greater gagefactor uncertainties than the 3M DP105-bonded sensing sheet The Araldite adhesive is much moreviscous than the 3M DP105 adhesive and did not spread as easily when bonding the sensor sheet to thealuminum beam resulting in minor air pockets under the sensor sheet With the 3M DP105 adhesiveit was much easier to spread an even layer on the aluminum beam and no air pockets were visiblelikely increasing apparent gage factor consistency both between and within sensor sheets

4 Conclusions

The aim of this paper was to determine a suitable adhesive for installation of a novel sensing sheetonto steel structural members subjected to cracking Previous research has shown that for sensingsheets installed using a stiff adhesive (eg Araldite) the cracking in steel would induce immediatefailure of individual sensors of the sheet To address this challenge three soft 3M glues were consideredDP100 DP105 and DP125

The 3M DP125 adhesive was eliminated because the set time was too long to be useful inpractical applications Strain transfer tests were performed using strain gages to assess the quality

Sensors 2018 18 1907 15 of 16

of strain transfer of the two other glues It was shown that while the Araldite adhesive transfersa large percentage of strain to the sensors bonding with both the 3M DP100 and DP105 providessufficient sensitivity to detect small strain changes even in the non-damage range No conclusiveresults were found to justify the preference between the DP100 or DP105 from the non-damage studyso crack opening damage tests were performed 48-hour and 72-hour cure tests showed that the DP105adhesive was able to detect strain changes for larger crack opening sizes compared to the DP100 so itwas deemed more suitable for damage detection (cracking) in mechanically non-degrading materials(eg metals)

A sensing sheet prototype manufactured at Princeton University was then tested using the 3MDP105 adhesive For each of the working individual sensors apparent gage factors were determinedby loading and unloading a thin aluminum beam with sand bottles using theoretical strain valuesunder Euler-Bernoulli beam theory Uncertainties in the apparent gage factor values were found forboth Araldite and 3M DP105-bonded sensor sheets As expected the apparent gage factors for theAraldite adhesive were higher than for the 3M DP105 adhesive We know that an approximate gagefactor for bonding a sensor sheet with 3M DP105 adhesive would allow one to use the sensor sheet toidentify real strain values on structures

The consistency between the strain transfer tests performed on regular strain gages and the testsperformed on sensing sheets confirm that DP105 can be used for the installation of sensing sheets onsteel structural members for crack detection and characterization purposes

Author Contributions BG NV and JES created the concept of sensing sheet MG designed and carried outthe tests and data analysis he is the lead author of the paper CW and LEA developed and manufacturedsensing sheet prototype and associated hardware

Acknowledgments This research was in part supported by the Princeton Institute for the Science andTechnology of Materials (PRISM) and in part by the USDOT-RITA UTC Program grant no DTRT12-G- UTC16enabled through the Center for Advanced Infrastructure and Transportation (CAIT) at Rutgers UniversityThe authors would like to thank Y Hu L Huang N Lin W Rieutort-Louis J Sanz-Robinson T Liu S Wagnerand J Vocaturo from Princeton University for their precious help Kaitlyn Kliever helped improve the contents ofthis paper

Conflicts of Interest The authors declare no conflict of interest

References

1 Yao Y Glišic B Reliable damage detection and localization using direct strain sensing In Proceedings ofthe Sixth International IAMBAS Conference on Bridge Maintenance Safety and Management Stresa Italy8ndash12 July 2012 pp 714ndash721

2 Yao Y Tung S-TE Glišic B Crack detection and characterization techniquesmdashAn overview Struct ControlHealth Monit 2014 21 1387ndash1413 [CrossRef]

3 Tung S-T Yao Y Glišic B Sensing Sheet The Sensitivity of Thin-Film Full- Bridge Strain Sensors for CrackDetection and Characterization Meas Sci Technol 2014 25 14 [CrossRef]

4 Glišic B Chen J Hubbell D Streicker Bridge A comparison between Bragg-gratings long-gaugestrain and temperature sensors and Brillouin scattering-based distributed strain and temperature sensorsIn Proceedings of the SPIE Smart StructuresNDE and Materials + Nondestructive Evaluation and HealthMonitoring San Diego CA USA 6ndash10 March 2011

5 Loh KJ Hou TC Lynch JP Kotov NA Carbon Nanotube Sensing Skins for Spatial Strain and ImpactDamage Identification J Nondestruct Eval 2009 28 9ndash25 [CrossRef]

6 Pyo S Loh KJ Hou TC Jarva E Lynch JP A wireless impedance analyzer for automated tomographicmapping of a nanoengineered sensing skin Smart Struct Syst 2011 8 139ndash155 [CrossRef]

7 Pour-Ghaz M Weiss J Application of Frequency Selective Circuits for Crack Detection in ConcreteElements J ASTM Int 2011 8 1ndash11

8 Schumacher T Thostenson ET Development of structural carbon nanotube-based sensing composites forconcrete structures J Intell Mater Syst Struct 2014 25 1331ndash1339 [CrossRef]

Sensors 2018 18 1907 9 of 16

Figure 9 Crack test setup

Based on the manufacturerrsquos specifications both adhesives achieve 80 of their full strengths ifcured for 20 minutes at 23 C and they achieve full strength if cured for 48 hours at 23 C Given thatthe temperature of the lab in which the tests were made was set to 17 C and could not be brought upto 23 C it was decided to perform one test before the full strength was achieved and one after the fullstrength was achieved For practical reasons the former was performed 24 hours after gluing and thelatter 72 hours after gluing (one day longer than specified time of 48 hours in order to account forlower curing temperature)

The following plots show the relationship between crack opening size and strain reading forthe 3M DP100 adhesive Both of these tests were performed with strain gages with a gage factor of201 In the first test (Figure 10) the adhesive was allowed to set for 24 hours and in the second testfor 72 hours (Figure 11) Time-dependent tests were performed in order to evaluate the adhesivesrsquoperformance with time and to establish a minimum cure time for future lab tests

Figure 10 DP100 crack opening vs strain after 24 hours

Sensors 2018 18 1907 10 of 16

Figure 11 DP100 crack opening vs strain after 72 hours

The time-dependent tests for the 3M DP100 adhesive showed very interesting results The testsdemonstrated that the bond strength does show time-dependence The 24-hour DP100 test showedreadings up to a crack opening of about 015 mm whereas the 72-hour test showed readings up toa 026 mm crack The result is important because it verifies that the adhesive performs better whengiven a few days to cure

Similar tests were performed for the 3M DP105 adhesive The results are shown inFigures 12 and 13 below

Figure 12 DP105 crack opening vs strain after 24 hours

Sensors 2018 18 1907 11 of 16

Figure 13 DP105 crack opening vs strain after 72 hours

The DP105 adhesive continued to read strain for crack openings up to 045 mm compared to about026 mm for the DP100 adhesive Taking into account that all other properties of the two adhesivesare similar the DP105 adhesive is considered the more suitable adhesive for the installation of sensingsheets that are supposed to detect and survive structural damage

3 Application to a Sensing Sheet Prototype

31 Objective

Once the appropriate adhesive for the sensing sheet was selected its application was testedon a sensing sheet prototype Given that manufacturing of a sensing sheet at the prototype stageis expensive it was decided to perform only non-damage strain transfer tests so that the sensing sheetwould not be damaged and could be used for future research related to hardware

The first step in implementing the sensing sheet is to determine appropriate gage factors of theindividual sensors and observe how these gage factors change depending on the adhesive used tobond the sheet The same testing setup as described in Section 23 was used The sensing sheet wasinstalled in the third quarter of the beam ie 1282 mm from the left support The load was applied1069 mm from the left support

The prototype of the sensing sheet is composed of eight individual sensors each composed offour resistors forming a full bridge Each individual sensor was independently tested by bonding thesensing sheet to an aluminum beam then loading and unloading the beam with sand bottles similarto the strain-transfer test described in the previous section In total two sensing sheets were testedthe first glued with the stiff Araldite adhesive for reference purposes and the second glued with thesoft DP105 glue Each loading step lasted 15 s and values were sampled at 1000 Hz 10 seconds intoeach loading step The beam was loaded for seven loading steps no load B1 B1 and B2 B1 and B2and B3 B1 and B2 B1 and no load giving a total testing time of 105 seconds At each loading step thechange in output voltage was recorded Using the change in output voltage the input voltage and ananalytical strain model the gage factor was determined for each sensor under each loading conditionFor each sensor identical tests were performed three times yielding three gage factors Preliminarytests included air and beam surface temperature measurements but no significant temperature changeswere found during the test duration Figures 14 and 15 show the test setup

Sensors 2018 18 1907 12 of 16

Figure 14 Sensing sheet and individual sensor numbering convention

Figure 15 Gage factor test setup The sensing sheet is bonded to the top surface of the aluminum beamand is connected to the rigid printed circuit board The temperature reading device is shown in thebottom left corner

32 Results

For each loading step for each individual sensor apparent gage factors were determined underidentical loading conditions and they are shown in Figure 16 Note that these gage factors are apparentand not true as they combine true gage factor with strain transfer quality The figure shows thatconsistent results are obtained for all individual sensors within the same sheet In addition the gagefactors for sensors attached using the DP105 adhesive are about 25 times less than those for the sensorsattached using Araldite (see Table 1) This result is consistent with strain transfer test results describedearlier in Section 2

As expected smaller loads yielded greater gage factor errors than large loads due to lower signalto noise ratios This created dispersion in the results Figure 17 plots the gage factor uncertaintymeasured as standard deviation for each load step Indices 1 and 5 represent the beam loaded withbottle B1 2 and 4 represent loading with bottles B1 and B2 and 3 represents loading with all three

Sensors 2018 18 1907 13 of 16

bottles Loading step 3 has the smallest gage factor uncertainty with a standard deviation of about006 for 3M DP105 and 013 for Araldite Loading step 5 has the greatest gage factor uncertaintywith a standard deviation of about 033 for 3M DP105 and 054 for Araldite As Figure 17 showsthe Araldite and 3M DP105 have similar gage factor uncertainty trends except the Aralditersquos gagefactor uncertainties are consistently greater than those for 3M DP105 The reason for this could be theimperfections and connection issues in the flexible to rigid interface where a zero-insertion-force (ZIF)connector is used Unwanted impedance in this connection can vary by insertion and introduces asmall amount of noise into the differential voltage which is used to calculate strain This also showsthe need for an integrated system implementation on a flexible substrate and the need to reduce thenumber of interfaces to rigid boards as mentioned in previous works [18ndash20]

Figure 16 Apparent gage factors determined on individual sensors of sensing sheet prototypes gluedwith Araldite and 3M DP105 adhesives

Figure 17 Apparent gage factor uncertainties for various load steps

Sensors 2018 18 1907 14 of 16

Table 1 Apparent gage factor summary for Araldite and 3M DP105 adhesives (microlowast = meanσlowast = standard deviation)

Araldite GF 3M DP105 GF

microlowast 135 054σlowast 0240 0128

33 Discussion

The apparent gage factor for Araldite is greater than the apparent gage factor for 3M DP105adhesive because the Araldite is less flexible and transfers more strain from the structure to the sensorStrain is proportional to change in output voltage and inversely proportional to gage factor so if ananalytical strain model is used to determine gage factor a stiff adhesive will result in a larger changein output voltage and the gage factor will be higher Obtained results are consistent with the results ofthe strain transfer test presented in Section 2

For the sensing sheet bonded with Araldite sensors 2 and 7 showed high voltage output noise inthe change in output voltage values and therefore did not yield reasonable gage factor values For thesensing sheet bonded with 3M DP105 sensor 2 had high noise and did not yield reasonable gagefactor values Because change in output voltage values for sensor 2 were unreasonable in both theAraldite and 3M DP105 tests it is likely that there is a design or manufacturing flaw in the sensingsheet itself or a soldering problem on the rigid printed circuit board (see Figure 15) In the case ofsensor 7 in the Araldite-bonded sensor sheet the high noise can likely be attributed to a manufacturingflaw on the sensor sheet itself

Figure 16 shows that there is a range in apparent gage factor values (dispersion) from sensorto sensor One reason the gage factors might vary from sensor to sensor has to do with a possiblevoltage drop from the sensor to the rigid printed circuit board and reading unit Because the wiresin the sensing sheet have high resistance sensors that are located farthest from the reading unit(sensors 0 and 4 for example) may show a lower voltage output change than sensors that are closerto the reading unit (sensors 3 and 8) Furthermore a lower voltage output change will result in alower gage factor Figure 16 shows generally lower gage factors for sensors 0 and 4 validating thishypothesis Sensors 1 and 5 show higher gage factors and exhibit similar behavior Sensor 7 doesnot follow the trend described which may be attributed to a manufacturing flaw Another reasonfor varying gage factors between sensors is that the adhesive layerrsquos thickness might vary across thesensor sheet A thinner adhesive layer will permit more strain to be transferred to the sensor and thegage factor will be higher Conversely a thicker adhesive layer will transfer less strain to the sensorand will result in a lower computed gage factor

Finally Figure 17 and Table 1 show that the Araldite-bonded sensing sheet yields greater gagefactor uncertainties than the 3M DP105-bonded sensing sheet The Araldite adhesive is much moreviscous than the 3M DP105 adhesive and did not spread as easily when bonding the sensor sheet to thealuminum beam resulting in minor air pockets under the sensor sheet With the 3M DP105 adhesiveit was much easier to spread an even layer on the aluminum beam and no air pockets were visiblelikely increasing apparent gage factor consistency both between and within sensor sheets

4 Conclusions

The aim of this paper was to determine a suitable adhesive for installation of a novel sensing sheetonto steel structural members subjected to cracking Previous research has shown that for sensingsheets installed using a stiff adhesive (eg Araldite) the cracking in steel would induce immediatefailure of individual sensors of the sheet To address this challenge three soft 3M glues were consideredDP100 DP105 and DP125

The 3M DP125 adhesive was eliminated because the set time was too long to be useful inpractical applications Strain transfer tests were performed using strain gages to assess the quality

Sensors 2018 18 1907 15 of 16

of strain transfer of the two other glues It was shown that while the Araldite adhesive transfersa large percentage of strain to the sensors bonding with both the 3M DP100 and DP105 providessufficient sensitivity to detect small strain changes even in the non-damage range No conclusiveresults were found to justify the preference between the DP100 or DP105 from the non-damage studyso crack opening damage tests were performed 48-hour and 72-hour cure tests showed that the DP105adhesive was able to detect strain changes for larger crack opening sizes compared to the DP100 so itwas deemed more suitable for damage detection (cracking) in mechanically non-degrading materials(eg metals)

A sensing sheet prototype manufactured at Princeton University was then tested using the 3MDP105 adhesive For each of the working individual sensors apparent gage factors were determinedby loading and unloading a thin aluminum beam with sand bottles using theoretical strain valuesunder Euler-Bernoulli beam theory Uncertainties in the apparent gage factor values were found forboth Araldite and 3M DP105-bonded sensor sheets As expected the apparent gage factors for theAraldite adhesive were higher than for the 3M DP105 adhesive We know that an approximate gagefactor for bonding a sensor sheet with 3M DP105 adhesive would allow one to use the sensor sheet toidentify real strain values on structures

The consistency between the strain transfer tests performed on regular strain gages and the testsperformed on sensing sheets confirm that DP105 can be used for the installation of sensing sheets onsteel structural members for crack detection and characterization purposes

Author Contributions BG NV and JES created the concept of sensing sheet MG designed and carried outthe tests and data analysis he is the lead author of the paper CW and LEA developed and manufacturedsensing sheet prototype and associated hardware

Acknowledgments This research was in part supported by the Princeton Institute for the Science andTechnology of Materials (PRISM) and in part by the USDOT-RITA UTC Program grant no DTRT12-G- UTC16enabled through the Center for Advanced Infrastructure and Transportation (CAIT) at Rutgers UniversityThe authors would like to thank Y Hu L Huang N Lin W Rieutort-Louis J Sanz-Robinson T Liu S Wagnerand J Vocaturo from Princeton University for their precious help Kaitlyn Kliever helped improve the contents ofthis paper

Conflicts of Interest The authors declare no conflict of interest

References

1 Yao Y Glišic B Reliable damage detection and localization using direct strain sensing In Proceedings ofthe Sixth International IAMBAS Conference on Bridge Maintenance Safety and Management Stresa Italy8ndash12 July 2012 pp 714ndash721

2 Yao Y Tung S-TE Glišic B Crack detection and characterization techniquesmdashAn overview Struct ControlHealth Monit 2014 21 1387ndash1413 [CrossRef]

3 Tung S-T Yao Y Glišic B Sensing Sheet The Sensitivity of Thin-Film Full- Bridge Strain Sensors for CrackDetection and Characterization Meas Sci Technol 2014 25 14 [CrossRef]

4 Glišic B Chen J Hubbell D Streicker Bridge A comparison between Bragg-gratings long-gaugestrain and temperature sensors and Brillouin scattering-based distributed strain and temperature sensorsIn Proceedings of the SPIE Smart StructuresNDE and Materials + Nondestructive Evaluation and HealthMonitoring San Diego CA USA 6ndash10 March 2011

5 Loh KJ Hou TC Lynch JP Kotov NA Carbon Nanotube Sensing Skins for Spatial Strain and ImpactDamage Identification J Nondestruct Eval 2009 28 9ndash25 [CrossRef]

6 Pyo S Loh KJ Hou TC Jarva E Lynch JP A wireless impedance analyzer for automated tomographicmapping of a nanoengineered sensing skin Smart Struct Syst 2011 8 139ndash155 [CrossRef]

7 Pour-Ghaz M Weiss J Application of Frequency Selective Circuits for Crack Detection in ConcreteElements J ASTM Int 2011 8 1ndash11

8 Schumacher T Thostenson ET Development of structural carbon nanotube-based sensing composites forconcrete structures J Intell Mater Syst Struct 2014 25 1331ndash1339 [CrossRef]

Sensors 2018 18 1907 10 of 16

Figure 11 DP100 crack opening vs strain after 72 hours

The time-dependent tests for the 3M DP100 adhesive showed very interesting results The testsdemonstrated that the bond strength does show time-dependence The 24-hour DP100 test showedreadings up to a crack opening of about 015 mm whereas the 72-hour test showed readings up toa 026 mm crack The result is important because it verifies that the adhesive performs better whengiven a few days to cure

Similar tests were performed for the 3M DP105 adhesive The results are shown inFigures 12 and 13 below

Figure 12 DP105 crack opening vs strain after 24 hours

Sensors 2018 18 1907 11 of 16

Figure 13 DP105 crack opening vs strain after 72 hours

The DP105 adhesive continued to read strain for crack openings up to 045 mm compared to about026 mm for the DP100 adhesive Taking into account that all other properties of the two adhesivesare similar the DP105 adhesive is considered the more suitable adhesive for the installation of sensingsheets that are supposed to detect and survive structural damage

3 Application to a Sensing Sheet Prototype

31 Objective

Once the appropriate adhesive for the sensing sheet was selected its application was testedon a sensing sheet prototype Given that manufacturing of a sensing sheet at the prototype stageis expensive it was decided to perform only non-damage strain transfer tests so that the sensing sheetwould not be damaged and could be used for future research related to hardware

The first step in implementing the sensing sheet is to determine appropriate gage factors of theindividual sensors and observe how these gage factors change depending on the adhesive used tobond the sheet The same testing setup as described in Section 23 was used The sensing sheet wasinstalled in the third quarter of the beam ie 1282 mm from the left support The load was applied1069 mm from the left support

The prototype of the sensing sheet is composed of eight individual sensors each composed offour resistors forming a full bridge Each individual sensor was independently tested by bonding thesensing sheet to an aluminum beam then loading and unloading the beam with sand bottles similarto the strain-transfer test described in the previous section In total two sensing sheets were testedthe first glued with the stiff Araldite adhesive for reference purposes and the second glued with thesoft DP105 glue Each loading step lasted 15 s and values were sampled at 1000 Hz 10 seconds intoeach loading step The beam was loaded for seven loading steps no load B1 B1 and B2 B1 and B2and B3 B1 and B2 B1 and no load giving a total testing time of 105 seconds At each loading step thechange in output voltage was recorded Using the change in output voltage the input voltage and ananalytical strain model the gage factor was determined for each sensor under each loading conditionFor each sensor identical tests were performed three times yielding three gage factors Preliminarytests included air and beam surface temperature measurements but no significant temperature changeswere found during the test duration Figures 14 and 15 show the test setup

Sensors 2018 18 1907 12 of 16

Figure 14 Sensing sheet and individual sensor numbering convention

Figure 15 Gage factor test setup The sensing sheet is bonded to the top surface of the aluminum beamand is connected to the rigid printed circuit board The temperature reading device is shown in thebottom left corner

32 Results

For each loading step for each individual sensor apparent gage factors were determined underidentical loading conditions and they are shown in Figure 16 Note that these gage factors are apparentand not true as they combine true gage factor with strain transfer quality The figure shows thatconsistent results are obtained for all individual sensors within the same sheet In addition the gagefactors for sensors attached using the DP105 adhesive are about 25 times less than those for the sensorsattached using Araldite (see Table 1) This result is consistent with strain transfer test results describedearlier in Section 2

As expected smaller loads yielded greater gage factor errors than large loads due to lower signalto noise ratios This created dispersion in the results Figure 17 plots the gage factor uncertaintymeasured as standard deviation for each load step Indices 1 and 5 represent the beam loaded withbottle B1 2 and 4 represent loading with bottles B1 and B2 and 3 represents loading with all three

Sensors 2018 18 1907 13 of 16

bottles Loading step 3 has the smallest gage factor uncertainty with a standard deviation of about006 for 3M DP105 and 013 for Araldite Loading step 5 has the greatest gage factor uncertaintywith a standard deviation of about 033 for 3M DP105 and 054 for Araldite As Figure 17 showsthe Araldite and 3M DP105 have similar gage factor uncertainty trends except the Aralditersquos gagefactor uncertainties are consistently greater than those for 3M DP105 The reason for this could be theimperfections and connection issues in the flexible to rigid interface where a zero-insertion-force (ZIF)connector is used Unwanted impedance in this connection can vary by insertion and introduces asmall amount of noise into the differential voltage which is used to calculate strain This also showsthe need for an integrated system implementation on a flexible substrate and the need to reduce thenumber of interfaces to rigid boards as mentioned in previous works [18ndash20]

Figure 16 Apparent gage factors determined on individual sensors of sensing sheet prototypes gluedwith Araldite and 3M DP105 adhesives

Figure 17 Apparent gage factor uncertainties for various load steps

Sensors 2018 18 1907 14 of 16

Table 1 Apparent gage factor summary for Araldite and 3M DP105 adhesives (microlowast = meanσlowast = standard deviation)

Araldite GF 3M DP105 GF

microlowast 135 054σlowast 0240 0128

33 Discussion

The apparent gage factor for Araldite is greater than the apparent gage factor for 3M DP105adhesive because the Araldite is less flexible and transfers more strain from the structure to the sensorStrain is proportional to change in output voltage and inversely proportional to gage factor so if ananalytical strain model is used to determine gage factor a stiff adhesive will result in a larger changein output voltage and the gage factor will be higher Obtained results are consistent with the results ofthe strain transfer test presented in Section 2

For the sensing sheet bonded with Araldite sensors 2 and 7 showed high voltage output noise inthe change in output voltage values and therefore did not yield reasonable gage factor values For thesensing sheet bonded with 3M DP105 sensor 2 had high noise and did not yield reasonable gagefactor values Because change in output voltage values for sensor 2 were unreasonable in both theAraldite and 3M DP105 tests it is likely that there is a design or manufacturing flaw in the sensingsheet itself or a soldering problem on the rigid printed circuit board (see Figure 15) In the case ofsensor 7 in the Araldite-bonded sensor sheet the high noise can likely be attributed to a manufacturingflaw on the sensor sheet itself

Figure 16 shows that there is a range in apparent gage factor values (dispersion) from sensorto sensor One reason the gage factors might vary from sensor to sensor has to do with a possiblevoltage drop from the sensor to the rigid printed circuit board and reading unit Because the wiresin the sensing sheet have high resistance sensors that are located farthest from the reading unit(sensors 0 and 4 for example) may show a lower voltage output change than sensors that are closerto the reading unit (sensors 3 and 8) Furthermore a lower voltage output change will result in alower gage factor Figure 16 shows generally lower gage factors for sensors 0 and 4 validating thishypothesis Sensors 1 and 5 show higher gage factors and exhibit similar behavior Sensor 7 doesnot follow the trend described which may be attributed to a manufacturing flaw Another reasonfor varying gage factors between sensors is that the adhesive layerrsquos thickness might vary across thesensor sheet A thinner adhesive layer will permit more strain to be transferred to the sensor and thegage factor will be higher Conversely a thicker adhesive layer will transfer less strain to the sensorand will result in a lower computed gage factor

Finally Figure 17 and Table 1 show that the Araldite-bonded sensing sheet yields greater gagefactor uncertainties than the 3M DP105-bonded sensing sheet The Araldite adhesive is much moreviscous than the 3M DP105 adhesive and did not spread as easily when bonding the sensor sheet to thealuminum beam resulting in minor air pockets under the sensor sheet With the 3M DP105 adhesiveit was much easier to spread an even layer on the aluminum beam and no air pockets were visiblelikely increasing apparent gage factor consistency both between and within sensor sheets

4 Conclusions

The aim of this paper was to determine a suitable adhesive for installation of a novel sensing sheetonto steel structural members subjected to cracking Previous research has shown that for sensingsheets installed using a stiff adhesive (eg Araldite) the cracking in steel would induce immediatefailure of individual sensors of the sheet To address this challenge three soft 3M glues were consideredDP100 DP105 and DP125

The 3M DP125 adhesive was eliminated because the set time was too long to be useful inpractical applications Strain transfer tests were performed using strain gages to assess the quality

Sensors 2018 18 1907 15 of 16

of strain transfer of the two other glues It was shown that while the Araldite adhesive transfersa large percentage of strain to the sensors bonding with both the 3M DP100 and DP105 providessufficient sensitivity to detect small strain changes even in the non-damage range No conclusiveresults were found to justify the preference between the DP100 or DP105 from the non-damage studyso crack opening damage tests were performed 48-hour and 72-hour cure tests showed that the DP105adhesive was able to detect strain changes for larger crack opening sizes compared to the DP100 so itwas deemed more suitable for damage detection (cracking) in mechanically non-degrading materials(eg metals)

A sensing sheet prototype manufactured at Princeton University was then tested using the 3MDP105 adhesive For each of the working individual sensors apparent gage factors were determinedby loading and unloading a thin aluminum beam with sand bottles using theoretical strain valuesunder Euler-Bernoulli beam theory Uncertainties in the apparent gage factor values were found forboth Araldite and 3M DP105-bonded sensor sheets As expected the apparent gage factors for theAraldite adhesive were higher than for the 3M DP105 adhesive We know that an approximate gagefactor for bonding a sensor sheet with 3M DP105 adhesive would allow one to use the sensor sheet toidentify real strain values on structures

The consistency between the strain transfer tests performed on regular strain gages and the testsperformed on sensing sheets confirm that DP105 can be used for the installation of sensing sheets onsteel structural members for crack detection and characterization purposes

Author Contributions BG NV and JES created the concept of sensing sheet MG designed and carried outthe tests and data analysis he is the lead author of the paper CW and LEA developed and manufacturedsensing sheet prototype and associated hardware

Acknowledgments This research was in part supported by the Princeton Institute for the Science andTechnology of Materials (PRISM) and in part by the USDOT-RITA UTC Program grant no DTRT12-G- UTC16enabled through the Center for Advanced Infrastructure and Transportation (CAIT) at Rutgers UniversityThe authors would like to thank Y Hu L Huang N Lin W Rieutort-Louis J Sanz-Robinson T Liu S Wagnerand J Vocaturo from Princeton University for their precious help Kaitlyn Kliever helped improve the contents ofthis paper

Conflicts of Interest The authors declare no conflict of interest

References

1 Yao Y Glišic B Reliable damage detection and localization using direct strain sensing In Proceedings ofthe Sixth International IAMBAS Conference on Bridge Maintenance Safety and Management Stresa Italy8ndash12 July 2012 pp 714ndash721

2 Yao Y Tung S-TE Glišic B Crack detection and characterization techniquesmdashAn overview Struct ControlHealth Monit 2014 21 1387ndash1413 [CrossRef]

3 Tung S-T Yao Y Glišic B Sensing Sheet The Sensitivity of Thin-Film Full- Bridge Strain Sensors for CrackDetection and Characterization Meas Sci Technol 2014 25 14 [CrossRef]

4 Glišic B Chen J Hubbell D Streicker Bridge A comparison between Bragg-gratings long-gaugestrain and temperature sensors and Brillouin scattering-based distributed strain and temperature sensorsIn Proceedings of the SPIE Smart StructuresNDE and Materials + Nondestructive Evaluation and HealthMonitoring San Diego CA USA 6ndash10 March 2011

5 Loh KJ Hou TC Lynch JP Kotov NA Carbon Nanotube Sensing Skins for Spatial Strain and ImpactDamage Identification J Nondestruct Eval 2009 28 9ndash25 [CrossRef]

6 Pyo S Loh KJ Hou TC Jarva E Lynch JP A wireless impedance analyzer for automated tomographicmapping of a nanoengineered sensing skin Smart Struct Syst 2011 8 139ndash155 [CrossRef]

7 Pour-Ghaz M Weiss J Application of Frequency Selective Circuits for Crack Detection in ConcreteElements J ASTM Int 2011 8 1ndash11

8 Schumacher T Thostenson ET Development of structural carbon nanotube-based sensing composites forconcrete structures J Intell Mater Syst Struct 2014 25 1331ndash1339 [CrossRef]

Sensors 2018 18 1907 11 of 16

Figure 13 DP105 crack opening vs strain after 72 hours

The DP105 adhesive continued to read strain for crack openings up to 045 mm compared to about026 mm for the DP100 adhesive Taking into account that all other properties of the two adhesivesare similar the DP105 adhesive is considered the more suitable adhesive for the installation of sensingsheets that are supposed to detect and survive structural damage

3 Application to a Sensing Sheet Prototype

31 Objective

Once the appropriate adhesive for the sensing sheet was selected its application was testedon a sensing sheet prototype Given that manufacturing of a sensing sheet at the prototype stageis expensive it was decided to perform only non-damage strain transfer tests so that the sensing sheetwould not be damaged and could be used for future research related to hardware

The first step in implementing the sensing sheet is to determine appropriate gage factors of theindividual sensors and observe how these gage factors change depending on the adhesive used tobond the sheet The same testing setup as described in Section 23 was used The sensing sheet wasinstalled in the third quarter of the beam ie 1282 mm from the left support The load was applied1069 mm from the left support

The prototype of the sensing sheet is composed of eight individual sensors each composed offour resistors forming a full bridge Each individual sensor was independently tested by bonding thesensing sheet to an aluminum beam then loading and unloading the beam with sand bottles similarto the strain-transfer test described in the previous section In total two sensing sheets were testedthe first glued with the stiff Araldite adhesive for reference purposes and the second glued with thesoft DP105 glue Each loading step lasted 15 s and values were sampled at 1000 Hz 10 seconds intoeach loading step The beam was loaded for seven loading steps no load B1 B1 and B2 B1 and B2and B3 B1 and B2 B1 and no load giving a total testing time of 105 seconds At each loading step thechange in output voltage was recorded Using the change in output voltage the input voltage and ananalytical strain model the gage factor was determined for each sensor under each loading conditionFor each sensor identical tests were performed three times yielding three gage factors Preliminarytests included air and beam surface temperature measurements but no significant temperature changeswere found during the test duration Figures 14 and 15 show the test setup

Sensors 2018 18 1907 12 of 16

Figure 14 Sensing sheet and individual sensor numbering convention

Figure 15 Gage factor test setup The sensing sheet is bonded to the top surface of the aluminum beamand is connected to the rigid printed circuit board The temperature reading device is shown in thebottom left corner

32 Results

For each loading step for each individual sensor apparent gage factors were determined underidentical loading conditions and they are shown in Figure 16 Note that these gage factors are apparentand not true as they combine true gage factor with strain transfer quality The figure shows thatconsistent results are obtained for all individual sensors within the same sheet In addition the gagefactors for sensors attached using the DP105 adhesive are about 25 times less than those for the sensorsattached using Araldite (see Table 1) This result is consistent with strain transfer test results describedearlier in Section 2

As expected smaller loads yielded greater gage factor errors than large loads due to lower signalto noise ratios This created dispersion in the results Figure 17 plots the gage factor uncertaintymeasured as standard deviation for each load step Indices 1 and 5 represent the beam loaded withbottle B1 2 and 4 represent loading with bottles B1 and B2 and 3 represents loading with all three

Sensors 2018 18 1907 13 of 16

bottles Loading step 3 has the smallest gage factor uncertainty with a standard deviation of about006 for 3M DP105 and 013 for Araldite Loading step 5 has the greatest gage factor uncertaintywith a standard deviation of about 033 for 3M DP105 and 054 for Araldite As Figure 17 showsthe Araldite and 3M DP105 have similar gage factor uncertainty trends except the Aralditersquos gagefactor uncertainties are consistently greater than those for 3M DP105 The reason for this could be theimperfections and connection issues in the flexible to rigid interface where a zero-insertion-force (ZIF)connector is used Unwanted impedance in this connection can vary by insertion and introduces asmall amount of noise into the differential voltage which is used to calculate strain This also showsthe need for an integrated system implementation on a flexible substrate and the need to reduce thenumber of interfaces to rigid boards as mentioned in previous works [18ndash20]

Figure 16 Apparent gage factors determined on individual sensors of sensing sheet prototypes gluedwith Araldite and 3M DP105 adhesives

Figure 17 Apparent gage factor uncertainties for various load steps

Sensors 2018 18 1907 14 of 16

Table 1 Apparent gage factor summary for Araldite and 3M DP105 adhesives (microlowast = meanσlowast = standard deviation)

Araldite GF 3M DP105 GF

microlowast 135 054σlowast 0240 0128

33 Discussion

The apparent gage factor for Araldite is greater than the apparent gage factor for 3M DP105adhesive because the Araldite is less flexible and transfers more strain from the structure to the sensorStrain is proportional to change in output voltage and inversely proportional to gage factor so if ananalytical strain model is used to determine gage factor a stiff adhesive will result in a larger changein output voltage and the gage factor will be higher Obtained results are consistent with the results ofthe strain transfer test presented in Section 2

For the sensing sheet bonded with Araldite sensors 2 and 7 showed high voltage output noise inthe change in output voltage values and therefore did not yield reasonable gage factor values For thesensing sheet bonded with 3M DP105 sensor 2 had high noise and did not yield reasonable gagefactor values Because change in output voltage values for sensor 2 were unreasonable in both theAraldite and 3M DP105 tests it is likely that there is a design or manufacturing flaw in the sensingsheet itself or a soldering problem on the rigid printed circuit board (see Figure 15) In the case ofsensor 7 in the Araldite-bonded sensor sheet the high noise can likely be attributed to a manufacturingflaw on the sensor sheet itself

Figure 16 shows that there is a range in apparent gage factor values (dispersion) from sensorto sensor One reason the gage factors might vary from sensor to sensor has to do with a possiblevoltage drop from the sensor to the rigid printed circuit board and reading unit Because the wiresin the sensing sheet have high resistance sensors that are located farthest from the reading unit(sensors 0 and 4 for example) may show a lower voltage output change than sensors that are closerto the reading unit (sensors 3 and 8) Furthermore a lower voltage output change will result in alower gage factor Figure 16 shows generally lower gage factors for sensors 0 and 4 validating thishypothesis Sensors 1 and 5 show higher gage factors and exhibit similar behavior Sensor 7 doesnot follow the trend described which may be attributed to a manufacturing flaw Another reasonfor varying gage factors between sensors is that the adhesive layerrsquos thickness might vary across thesensor sheet A thinner adhesive layer will permit more strain to be transferred to the sensor and thegage factor will be higher Conversely a thicker adhesive layer will transfer less strain to the sensorand will result in a lower computed gage factor

Finally Figure 17 and Table 1 show that the Araldite-bonded sensing sheet yields greater gagefactor uncertainties than the 3M DP105-bonded sensing sheet The Araldite adhesive is much moreviscous than the 3M DP105 adhesive and did not spread as easily when bonding the sensor sheet to thealuminum beam resulting in minor air pockets under the sensor sheet With the 3M DP105 adhesiveit was much easier to spread an even layer on the aluminum beam and no air pockets were visiblelikely increasing apparent gage factor consistency both between and within sensor sheets

4 Conclusions

The aim of this paper was to determine a suitable adhesive for installation of a novel sensing sheetonto steel structural members subjected to cracking Previous research has shown that for sensingsheets installed using a stiff adhesive (eg Araldite) the cracking in steel would induce immediatefailure of individual sensors of the sheet To address this challenge three soft 3M glues were consideredDP100 DP105 and DP125

The 3M DP125 adhesive was eliminated because the set time was too long to be useful inpractical applications Strain transfer tests were performed using strain gages to assess the quality

Sensors 2018 18 1907 15 of 16

of strain transfer of the two other glues It was shown that while the Araldite adhesive transfersa large percentage of strain to the sensors bonding with both the 3M DP100 and DP105 providessufficient sensitivity to detect small strain changes even in the non-damage range No conclusiveresults were found to justify the preference between the DP100 or DP105 from the non-damage studyso crack opening damage tests were performed 48-hour and 72-hour cure tests showed that the DP105adhesive was able to detect strain changes for larger crack opening sizes compared to the DP100 so itwas deemed more suitable for damage detection (cracking) in mechanically non-degrading materials(eg metals)

A sensing sheet prototype manufactured at Princeton University was then tested using the 3MDP105 adhesive For each of the working individual sensors apparent gage factors were determinedby loading and unloading a thin aluminum beam with sand bottles using theoretical strain valuesunder Euler-Bernoulli beam theory Uncertainties in the apparent gage factor values were found forboth Araldite and 3M DP105-bonded sensor sheets As expected the apparent gage factors for theAraldite adhesive were higher than for the 3M DP105 adhesive We know that an approximate gagefactor for bonding a sensor sheet with 3M DP105 adhesive would allow one to use the sensor sheet toidentify real strain values on structures

The consistency between the strain transfer tests performed on regular strain gages and the testsperformed on sensing sheets confirm that DP105 can be used for the installation of sensing sheets onsteel structural members for crack detection and characterization purposes

Author Contributions BG NV and JES created the concept of sensing sheet MG designed and carried outthe tests and data analysis he is the lead author of the paper CW and LEA developed and manufacturedsensing sheet prototype and associated hardware

Acknowledgments This research was in part supported by the Princeton Institute for the Science andTechnology of Materials (PRISM) and in part by the USDOT-RITA UTC Program grant no DTRT12-G- UTC16enabled through the Center for Advanced Infrastructure and Transportation (CAIT) at Rutgers UniversityThe authors would like to thank Y Hu L Huang N Lin W Rieutort-Louis J Sanz-Robinson T Liu S Wagnerand J Vocaturo from Princeton University for their precious help Kaitlyn Kliever helped improve the contents ofthis paper

Conflicts of Interest The authors declare no conflict of interest

References

1 Yao Y Glišic B Reliable damage detection and localization using direct strain sensing In Proceedings ofthe Sixth International IAMBAS Conference on Bridge Maintenance Safety and Management Stresa Italy8ndash12 July 2012 pp 714ndash721

2 Yao Y Tung S-TE Glišic B Crack detection and characterization techniquesmdashAn overview Struct ControlHealth Monit 2014 21 1387ndash1413 [CrossRef]

3 Tung S-T Yao Y Glišic B Sensing Sheet The Sensitivity of Thin-Film Full- Bridge Strain Sensors for CrackDetection and Characterization Meas Sci Technol 2014 25 14 [CrossRef]

4 Glišic B Chen J Hubbell D Streicker Bridge A comparison between Bragg-gratings long-gaugestrain and temperature sensors and Brillouin scattering-based distributed strain and temperature sensorsIn Proceedings of the SPIE Smart StructuresNDE and Materials + Nondestructive Evaluation and HealthMonitoring San Diego CA USA 6ndash10 March 2011

5 Loh KJ Hou TC Lynch JP Kotov NA Carbon Nanotube Sensing Skins for Spatial Strain and ImpactDamage Identification J Nondestruct Eval 2009 28 9ndash25 [CrossRef]

6 Pyo S Loh KJ Hou TC Jarva E Lynch JP A wireless impedance analyzer for automated tomographicmapping of a nanoengineered sensing skin Smart Struct Syst 2011 8 139ndash155 [CrossRef]

7 Pour-Ghaz M Weiss J Application of Frequency Selective Circuits for Crack Detection in ConcreteElements J ASTM Int 2011 8 1ndash11

8 Schumacher T Thostenson ET Development of structural carbon nanotube-based sensing composites forconcrete structures J Intell Mater Syst Struct 2014 25 1331ndash1339 [CrossRef]

Sensors 2018 18 1907 12 of 16

Figure 14 Sensing sheet and individual sensor numbering convention

Figure 15 Gage factor test setup The sensing sheet is bonded to the top surface of the aluminum beamand is connected to the rigid printed circuit board The temperature reading device is shown in thebottom left corner

32 Results

For each loading step for each individual sensor apparent gage factors were determined underidentical loading conditions and they are shown in Figure 16 Note that these gage factors are apparentand not true as they combine true gage factor with strain transfer quality The figure shows thatconsistent results are obtained for all individual sensors within the same sheet In addition the gagefactors for sensors attached using the DP105 adhesive are about 25 times less than those for the sensorsattached using Araldite (see Table 1) This result is consistent with strain transfer test results describedearlier in Section 2

As expected smaller loads yielded greater gage factor errors than large loads due to lower signalto noise ratios This created dispersion in the results Figure 17 plots the gage factor uncertaintymeasured as standard deviation for each load step Indices 1 and 5 represent the beam loaded withbottle B1 2 and 4 represent loading with bottles B1 and B2 and 3 represents loading with all three

Sensors 2018 18 1907 13 of 16

bottles Loading step 3 has the smallest gage factor uncertainty with a standard deviation of about006 for 3M DP105 and 013 for Araldite Loading step 5 has the greatest gage factor uncertaintywith a standard deviation of about 033 for 3M DP105 and 054 for Araldite As Figure 17 showsthe Araldite and 3M DP105 have similar gage factor uncertainty trends except the Aralditersquos gagefactor uncertainties are consistently greater than those for 3M DP105 The reason for this could be theimperfections and connection issues in the flexible to rigid interface where a zero-insertion-force (ZIF)connector is used Unwanted impedance in this connection can vary by insertion and introduces asmall amount of noise into the differential voltage which is used to calculate strain This also showsthe need for an integrated system implementation on a flexible substrate and the need to reduce thenumber of interfaces to rigid boards as mentioned in previous works [18ndash20]

Figure 16 Apparent gage factors determined on individual sensors of sensing sheet prototypes gluedwith Araldite and 3M DP105 adhesives

Figure 17 Apparent gage factor uncertainties for various load steps

Sensors 2018 18 1907 14 of 16

Table 1 Apparent gage factor summary for Araldite and 3M DP105 adhesives (microlowast = meanσlowast = standard deviation)

Araldite GF 3M DP105 GF

microlowast 135 054σlowast 0240 0128

33 Discussion

The apparent gage factor for Araldite is greater than the apparent gage factor for 3M DP105adhesive because the Araldite is less flexible and transfers more strain from the structure to the sensorStrain is proportional to change in output voltage and inversely proportional to gage factor so if ananalytical strain model is used to determine gage factor a stiff adhesive will result in a larger changein output voltage and the gage factor will be higher Obtained results are consistent with the results ofthe strain transfer test presented in Section 2

For the sensing sheet bonded with Araldite sensors 2 and 7 showed high voltage output noise inthe change in output voltage values and therefore did not yield reasonable gage factor values For thesensing sheet bonded with 3M DP105 sensor 2 had high noise and did not yield reasonable gagefactor values Because change in output voltage values for sensor 2 were unreasonable in both theAraldite and 3M DP105 tests it is likely that there is a design or manufacturing flaw in the sensingsheet itself or a soldering problem on the rigid printed circuit board (see Figure 15) In the case ofsensor 7 in the Araldite-bonded sensor sheet the high noise can likely be attributed to a manufacturingflaw on the sensor sheet itself

Figure 16 shows that there is a range in apparent gage factor values (dispersion) from sensorto sensor One reason the gage factors might vary from sensor to sensor has to do with a possiblevoltage drop from the sensor to the rigid printed circuit board and reading unit Because the wiresin the sensing sheet have high resistance sensors that are located farthest from the reading unit(sensors 0 and 4 for example) may show a lower voltage output change than sensors that are closerto the reading unit (sensors 3 and 8) Furthermore a lower voltage output change will result in alower gage factor Figure 16 shows generally lower gage factors for sensors 0 and 4 validating thishypothesis Sensors 1 and 5 show higher gage factors and exhibit similar behavior Sensor 7 doesnot follow the trend described which may be attributed to a manufacturing flaw Another reasonfor varying gage factors between sensors is that the adhesive layerrsquos thickness might vary across thesensor sheet A thinner adhesive layer will permit more strain to be transferred to the sensor and thegage factor will be higher Conversely a thicker adhesive layer will transfer less strain to the sensorand will result in a lower computed gage factor

Finally Figure 17 and Table 1 show that the Araldite-bonded sensing sheet yields greater gagefactor uncertainties than the 3M DP105-bonded sensing sheet The Araldite adhesive is much moreviscous than the 3M DP105 adhesive and did not spread as easily when bonding the sensor sheet to thealuminum beam resulting in minor air pockets under the sensor sheet With the 3M DP105 adhesiveit was much easier to spread an even layer on the aluminum beam and no air pockets were visiblelikely increasing apparent gage factor consistency both between and within sensor sheets

4 Conclusions

The aim of this paper was to determine a suitable adhesive for installation of a novel sensing sheetonto steel structural members subjected to cracking Previous research has shown that for sensingsheets installed using a stiff adhesive (eg Araldite) the cracking in steel would induce immediatefailure of individual sensors of the sheet To address this challenge three soft 3M glues were consideredDP100 DP105 and DP125

The 3M DP125 adhesive was eliminated because the set time was too long to be useful inpractical applications Strain transfer tests were performed using strain gages to assess the quality

Sensors 2018 18 1907 15 of 16

of strain transfer of the two other glues It was shown that while the Araldite adhesive transfersa large percentage of strain to the sensors bonding with both the 3M DP100 and DP105 providessufficient sensitivity to detect small strain changes even in the non-damage range No conclusiveresults were found to justify the preference between the DP100 or DP105 from the non-damage studyso crack opening damage tests were performed 48-hour and 72-hour cure tests showed that the DP105adhesive was able to detect strain changes for larger crack opening sizes compared to the DP100 so itwas deemed more suitable for damage detection (cracking) in mechanically non-degrading materials(eg metals)

A sensing sheet prototype manufactured at Princeton University was then tested using the 3MDP105 adhesive For each of the working individual sensors apparent gage factors were determinedby loading and unloading a thin aluminum beam with sand bottles using theoretical strain valuesunder Euler-Bernoulli beam theory Uncertainties in the apparent gage factor values were found forboth Araldite and 3M DP105-bonded sensor sheets As expected the apparent gage factors for theAraldite adhesive were higher than for the 3M DP105 adhesive We know that an approximate gagefactor for bonding a sensor sheet with 3M DP105 adhesive would allow one to use the sensor sheet toidentify real strain values on structures

The consistency between the strain transfer tests performed on regular strain gages and the testsperformed on sensing sheets confirm that DP105 can be used for the installation of sensing sheets onsteel structural members for crack detection and characterization purposes

Author Contributions BG NV and JES created the concept of sensing sheet MG designed and carried outthe tests and data analysis he is the lead author of the paper CW and LEA developed and manufacturedsensing sheet prototype and associated hardware

Acknowledgments This research was in part supported by the Princeton Institute for the Science andTechnology of Materials (PRISM) and in part by the USDOT-RITA UTC Program grant no DTRT12-G- UTC16enabled through the Center for Advanced Infrastructure and Transportation (CAIT) at Rutgers UniversityThe authors would like to thank Y Hu L Huang N Lin W Rieutort-Louis J Sanz-Robinson T Liu S Wagnerand J Vocaturo from Princeton University for their precious help Kaitlyn Kliever helped improve the contents ofthis paper

Conflicts of Interest The authors declare no conflict of interest

References

1 Yao Y Glišic B Reliable damage detection and localization using direct strain sensing In Proceedings ofthe Sixth International IAMBAS Conference on Bridge Maintenance Safety and Management Stresa Italy8ndash12 July 2012 pp 714ndash721

2 Yao Y Tung S-TE Glišic B Crack detection and characterization techniquesmdashAn overview Struct ControlHealth Monit 2014 21 1387ndash1413 [CrossRef]

3 Tung S-T Yao Y Glišic B Sensing Sheet The Sensitivity of Thin-Film Full- Bridge Strain Sensors for CrackDetection and Characterization Meas Sci Technol 2014 25 14 [CrossRef]

4 Glišic B Chen J Hubbell D Streicker Bridge A comparison between Bragg-gratings long-gaugestrain and temperature sensors and Brillouin scattering-based distributed strain and temperature sensorsIn Proceedings of the SPIE Smart StructuresNDE and Materials + Nondestructive Evaluation and HealthMonitoring San Diego CA USA 6ndash10 March 2011

5 Loh KJ Hou TC Lynch JP Kotov NA Carbon Nanotube Sensing Skins for Spatial Strain and ImpactDamage Identification J Nondestruct Eval 2009 28 9ndash25 [CrossRef]

6 Pyo S Loh KJ Hou TC Jarva E Lynch JP A wireless impedance analyzer for automated tomographicmapping of a nanoengineered sensing skin Smart Struct Syst 2011 8 139ndash155 [CrossRef]

7 Pour-Ghaz M Weiss J Application of Frequency Selective Circuits for Crack Detection in ConcreteElements J ASTM Int 2011 8 1ndash11

8 Schumacher T Thostenson ET Development of structural carbon nanotube-based sensing composites forconcrete structures J Intell Mater Syst Struct 2014 25 1331ndash1339 [CrossRef]

Sensors 2018 18 1907 13 of 16

bottles Loading step 3 has the smallest gage factor uncertainty with a standard deviation of about006 for 3M DP105 and 013 for Araldite Loading step 5 has the greatest gage factor uncertaintywith a standard deviation of about 033 for 3M DP105 and 054 for Araldite As Figure 17 showsthe Araldite and 3M DP105 have similar gage factor uncertainty trends except the Aralditersquos gagefactor uncertainties are consistently greater than those for 3M DP105 The reason for this could be theimperfections and connection issues in the flexible to rigid interface where a zero-insertion-force (ZIF)connector is used Unwanted impedance in this connection can vary by insertion and introduces asmall amount of noise into the differential voltage which is used to calculate strain This also showsthe need for an integrated system implementation on a flexible substrate and the need to reduce thenumber of interfaces to rigid boards as mentioned in previous works [18ndash20]

Figure 16 Apparent gage factors determined on individual sensors of sensing sheet prototypes gluedwith Araldite and 3M DP105 adhesives

Figure 17 Apparent gage factor uncertainties for various load steps

Sensors 2018 18 1907 14 of 16

Table 1 Apparent gage factor summary for Araldite and 3M DP105 adhesives (microlowast = meanσlowast = standard deviation)

Araldite GF 3M DP105 GF

microlowast 135 054σlowast 0240 0128

33 Discussion

The apparent gage factor for Araldite is greater than the apparent gage factor for 3M DP105adhesive because the Araldite is less flexible and transfers more strain from the structure to the sensorStrain is proportional to change in output voltage and inversely proportional to gage factor so if ananalytical strain model is used to determine gage factor a stiff adhesive will result in a larger changein output voltage and the gage factor will be higher Obtained results are consistent with the results ofthe strain transfer test presented in Section 2

For the sensing sheet bonded with Araldite sensors 2 and 7 showed high voltage output noise inthe change in output voltage values and therefore did not yield reasonable gage factor values For thesensing sheet bonded with 3M DP105 sensor 2 had high noise and did not yield reasonable gagefactor values Because change in output voltage values for sensor 2 were unreasonable in both theAraldite and 3M DP105 tests it is likely that there is a design or manufacturing flaw in the sensingsheet itself or a soldering problem on the rigid printed circuit board (see Figure 15) In the case ofsensor 7 in the Araldite-bonded sensor sheet the high noise can likely be attributed to a manufacturingflaw on the sensor sheet itself

Figure 16 shows that there is a range in apparent gage factor values (dispersion) from sensorto sensor One reason the gage factors might vary from sensor to sensor has to do with a possiblevoltage drop from the sensor to the rigid printed circuit board and reading unit Because the wiresin the sensing sheet have high resistance sensors that are located farthest from the reading unit(sensors 0 and 4 for example) may show a lower voltage output change than sensors that are closerto the reading unit (sensors 3 and 8) Furthermore a lower voltage output change will result in alower gage factor Figure 16 shows generally lower gage factors for sensors 0 and 4 validating thishypothesis Sensors 1 and 5 show higher gage factors and exhibit similar behavior Sensor 7 doesnot follow the trend described which may be attributed to a manufacturing flaw Another reasonfor varying gage factors between sensors is that the adhesive layerrsquos thickness might vary across thesensor sheet A thinner adhesive layer will permit more strain to be transferred to the sensor and thegage factor will be higher Conversely a thicker adhesive layer will transfer less strain to the sensorand will result in a lower computed gage factor

Finally Figure 17 and Table 1 show that the Araldite-bonded sensing sheet yields greater gagefactor uncertainties than the 3M DP105-bonded sensing sheet The Araldite adhesive is much moreviscous than the 3M DP105 adhesive and did not spread as easily when bonding the sensor sheet to thealuminum beam resulting in minor air pockets under the sensor sheet With the 3M DP105 adhesiveit was much easier to spread an even layer on the aluminum beam and no air pockets were visiblelikely increasing apparent gage factor consistency both between and within sensor sheets

4 Conclusions

The aim of this paper was to determine a suitable adhesive for installation of a novel sensing sheetonto steel structural members subjected to cracking Previous research has shown that for sensingsheets installed using a stiff adhesive (eg Araldite) the cracking in steel would induce immediatefailure of individual sensors of the sheet To address this challenge three soft 3M glues were consideredDP100 DP105 and DP125

The 3M DP125 adhesive was eliminated because the set time was too long to be useful inpractical applications Strain transfer tests were performed using strain gages to assess the quality

Sensors 2018 18 1907 15 of 16

of strain transfer of the two other glues It was shown that while the Araldite adhesive transfersa large percentage of strain to the sensors bonding with both the 3M DP100 and DP105 providessufficient sensitivity to detect small strain changes even in the non-damage range No conclusiveresults were found to justify the preference between the DP100 or DP105 from the non-damage studyso crack opening damage tests were performed 48-hour and 72-hour cure tests showed that the DP105adhesive was able to detect strain changes for larger crack opening sizes compared to the DP100 so itwas deemed more suitable for damage detection (cracking) in mechanically non-degrading materials(eg metals)

A sensing sheet prototype manufactured at Princeton University was then tested using the 3MDP105 adhesive For each of the working individual sensors apparent gage factors were determinedby loading and unloading a thin aluminum beam with sand bottles using theoretical strain valuesunder Euler-Bernoulli beam theory Uncertainties in the apparent gage factor values were found forboth Araldite and 3M DP105-bonded sensor sheets As expected the apparent gage factors for theAraldite adhesive were higher than for the 3M DP105 adhesive We know that an approximate gagefactor for bonding a sensor sheet with 3M DP105 adhesive would allow one to use the sensor sheet toidentify real strain values on structures

The consistency between the strain transfer tests performed on regular strain gages and the testsperformed on sensing sheets confirm that DP105 can be used for the installation of sensing sheets onsteel structural members for crack detection and characterization purposes

Author Contributions BG NV and JES created the concept of sensing sheet MG designed and carried outthe tests and data analysis he is the lead author of the paper CW and LEA developed and manufacturedsensing sheet prototype and associated hardware

Acknowledgments This research was in part supported by the Princeton Institute for the Science andTechnology of Materials (PRISM) and in part by the USDOT-RITA UTC Program grant no DTRT12-G- UTC16enabled through the Center for Advanced Infrastructure and Transportation (CAIT) at Rutgers UniversityThe authors would like to thank Y Hu L Huang N Lin W Rieutort-Louis J Sanz-Robinson T Liu S Wagnerand J Vocaturo from Princeton University for their precious help Kaitlyn Kliever helped improve the contents ofthis paper

Conflicts of Interest The authors declare no conflict of interest

References

1 Yao Y Glišic B Reliable damage detection and localization using direct strain sensing In Proceedings ofthe Sixth International IAMBAS Conference on Bridge Maintenance Safety and Management Stresa Italy8ndash12 July 2012 pp 714ndash721

2 Yao Y Tung S-TE Glišic B Crack detection and characterization techniquesmdashAn overview Struct ControlHealth Monit 2014 21 1387ndash1413 [CrossRef]

3 Tung S-T Yao Y Glišic B Sensing Sheet The Sensitivity of Thin-Film Full- Bridge Strain Sensors for CrackDetection and Characterization Meas Sci Technol 2014 25 14 [CrossRef]

4 Glišic B Chen J Hubbell D Streicker Bridge A comparison between Bragg-gratings long-gaugestrain and temperature sensors and Brillouin scattering-based distributed strain and temperature sensorsIn Proceedings of the SPIE Smart StructuresNDE and Materials + Nondestructive Evaluation and HealthMonitoring San Diego CA USA 6ndash10 March 2011

5 Loh KJ Hou TC Lynch JP Kotov NA Carbon Nanotube Sensing Skins for Spatial Strain and ImpactDamage Identification J Nondestruct Eval 2009 28 9ndash25 [CrossRef]

6 Pyo S Loh KJ Hou TC Jarva E Lynch JP A wireless impedance analyzer for automated tomographicmapping of a nanoengineered sensing skin Smart Struct Syst 2011 8 139ndash155 [CrossRef]

7 Pour-Ghaz M Weiss J Application of Frequency Selective Circuits for Crack Detection in ConcreteElements J ASTM Int 2011 8 1ndash11

8 Schumacher T Thostenson ET Development of structural carbon nanotube-based sensing composites forconcrete structures J Intell Mater Syst Struct 2014 25 1331ndash1339 [CrossRef]

Sensors 2018 18 1907 14 of 16

Table 1 Apparent gage factor summary for Araldite and 3M DP105 adhesives (microlowast = meanσlowast = standard deviation)

Araldite GF 3M DP105 GF

microlowast 135 054σlowast 0240 0128

33 Discussion

The apparent gage factor for Araldite is greater than the apparent gage factor for 3M DP105adhesive because the Araldite is less flexible and transfers more strain from the structure to the sensorStrain is proportional to change in output voltage and inversely proportional to gage factor so if ananalytical strain model is used to determine gage factor a stiff adhesive will result in a larger changein output voltage and the gage factor will be higher Obtained results are consistent with the results ofthe strain transfer test presented in Section 2

For the sensing sheet bonded with Araldite sensors 2 and 7 showed high voltage output noise inthe change in output voltage values and therefore did not yield reasonable gage factor values For thesensing sheet bonded with 3M DP105 sensor 2 had high noise and did not yield reasonable gagefactor values Because change in output voltage values for sensor 2 were unreasonable in both theAraldite and 3M DP105 tests it is likely that there is a design or manufacturing flaw in the sensingsheet itself or a soldering problem on the rigid printed circuit board (see Figure 15) In the case ofsensor 7 in the Araldite-bonded sensor sheet the high noise can likely be attributed to a manufacturingflaw on the sensor sheet itself

Figure 16 shows that there is a range in apparent gage factor values (dispersion) from sensorto sensor One reason the gage factors might vary from sensor to sensor has to do with a possiblevoltage drop from the sensor to the rigid printed circuit board and reading unit Because the wiresin the sensing sheet have high resistance sensors that are located farthest from the reading unit(sensors 0 and 4 for example) may show a lower voltage output change than sensors that are closerto the reading unit (sensors 3 and 8) Furthermore a lower voltage output change will result in alower gage factor Figure 16 shows generally lower gage factors for sensors 0 and 4 validating thishypothesis Sensors 1 and 5 show higher gage factors and exhibit similar behavior Sensor 7 doesnot follow the trend described which may be attributed to a manufacturing flaw Another reasonfor varying gage factors between sensors is that the adhesive layerrsquos thickness might vary across thesensor sheet A thinner adhesive layer will permit more strain to be transferred to the sensor and thegage factor will be higher Conversely a thicker adhesive layer will transfer less strain to the sensorand will result in a lower computed gage factor

Finally Figure 17 and Table 1 show that the Araldite-bonded sensing sheet yields greater gagefactor uncertainties than the 3M DP105-bonded sensing sheet The Araldite adhesive is much moreviscous than the 3M DP105 adhesive and did not spread as easily when bonding the sensor sheet to thealuminum beam resulting in minor air pockets under the sensor sheet With the 3M DP105 adhesiveit was much easier to spread an even layer on the aluminum beam and no air pockets were visiblelikely increasing apparent gage factor consistency both between and within sensor sheets

4 Conclusions

The aim of this paper was to determine a suitable adhesive for installation of a novel sensing sheetonto steel structural members subjected to cracking Previous research has shown that for sensingsheets installed using a stiff adhesive (eg Araldite) the cracking in steel would induce immediatefailure of individual sensors of the sheet To address this challenge three soft 3M glues were consideredDP100 DP105 and DP125

The 3M DP125 adhesive was eliminated because the set time was too long to be useful inpractical applications Strain transfer tests were performed using strain gages to assess the quality

Sensors 2018 18 1907 15 of 16

of strain transfer of the two other glues It was shown that while the Araldite adhesive transfersa large percentage of strain to the sensors bonding with both the 3M DP100 and DP105 providessufficient sensitivity to detect small strain changes even in the non-damage range No conclusiveresults were found to justify the preference between the DP100 or DP105 from the non-damage studyso crack opening damage tests were performed 48-hour and 72-hour cure tests showed that the DP105adhesive was able to detect strain changes for larger crack opening sizes compared to the DP100 so itwas deemed more suitable for damage detection (cracking) in mechanically non-degrading materials(eg metals)

A sensing sheet prototype manufactured at Princeton University was then tested using the 3MDP105 adhesive For each of the working individual sensors apparent gage factors were determinedby loading and unloading a thin aluminum beam with sand bottles using theoretical strain valuesunder Euler-Bernoulli beam theory Uncertainties in the apparent gage factor values were found forboth Araldite and 3M DP105-bonded sensor sheets As expected the apparent gage factors for theAraldite adhesive were higher than for the 3M DP105 adhesive We know that an approximate gagefactor for bonding a sensor sheet with 3M DP105 adhesive would allow one to use the sensor sheet toidentify real strain values on structures

The consistency between the strain transfer tests performed on regular strain gages and the testsperformed on sensing sheets confirm that DP105 can be used for the installation of sensing sheets onsteel structural members for crack detection and characterization purposes

Author Contributions BG NV and JES created the concept of sensing sheet MG designed and carried outthe tests and data analysis he is the lead author of the paper CW and LEA developed and manufacturedsensing sheet prototype and associated hardware

Acknowledgments This research was in part supported by the Princeton Institute for the Science andTechnology of Materials (PRISM) and in part by the USDOT-RITA UTC Program grant no DTRT12-G- UTC16enabled through the Center for Advanced Infrastructure and Transportation (CAIT) at Rutgers UniversityThe authors would like to thank Y Hu L Huang N Lin W Rieutort-Louis J Sanz-Robinson T Liu S Wagnerand J Vocaturo from Princeton University for their precious help Kaitlyn Kliever helped improve the contents ofthis paper

Conflicts of Interest The authors declare no conflict of interest

References

1 Yao Y Glišic B Reliable damage detection and localization using direct strain sensing In Proceedings ofthe Sixth International IAMBAS Conference on Bridge Maintenance Safety and Management Stresa Italy8ndash12 July 2012 pp 714ndash721

2 Yao Y Tung S-TE Glišic B Crack detection and characterization techniquesmdashAn overview Struct ControlHealth Monit 2014 21 1387ndash1413 [CrossRef]

3 Tung S-T Yao Y Glišic B Sensing Sheet The Sensitivity of Thin-Film Full- Bridge Strain Sensors for CrackDetection and Characterization Meas Sci Technol 2014 25 14 [CrossRef]

4 Glišic B Chen J Hubbell D Streicker Bridge A comparison between Bragg-gratings long-gaugestrain and temperature sensors and Brillouin scattering-based distributed strain and temperature sensorsIn Proceedings of the SPIE Smart StructuresNDE and Materials + Nondestructive Evaluation and HealthMonitoring San Diego CA USA 6ndash10 March 2011

5 Loh KJ Hou TC Lynch JP Kotov NA Carbon Nanotube Sensing Skins for Spatial Strain and ImpactDamage Identification J Nondestruct Eval 2009 28 9ndash25 [CrossRef]

6 Pyo S Loh KJ Hou TC Jarva E Lynch JP A wireless impedance analyzer for automated tomographicmapping of a nanoengineered sensing skin Smart Struct Syst 2011 8 139ndash155 [CrossRef]

7 Pour-Ghaz M Weiss J Application of Frequency Selective Circuits for Crack Detection in ConcreteElements J ASTM Int 2011 8 1ndash11

8 Schumacher T Thostenson ET Development of structural carbon nanotube-based sensing composites forconcrete structures J Intell Mater Syst Struct 2014 25 1331ndash1339 [CrossRef]

Sensors 2018 18 1907 15 of 16

of strain transfer of the two other glues It was shown that while the Araldite adhesive transfersa large percentage of strain to the sensors bonding with both the 3M DP100 and DP105 providessufficient sensitivity to detect small strain changes even in the non-damage range No conclusiveresults were found to justify the preference between the DP100 or DP105 from the non-damage studyso crack opening damage tests were performed 48-hour and 72-hour cure tests showed that the DP105adhesive was able to detect strain changes for larger crack opening sizes compared to the DP100 so itwas deemed more suitable for damage detection (cracking) in mechanically non-degrading materials(eg metals)

A sensing sheet prototype manufactured at Princeton University was then tested using the 3MDP105 adhesive For each of the working individual sensors apparent gage factors were determinedby loading and unloading a thin aluminum beam with sand bottles using theoretical strain valuesunder Euler-Bernoulli beam theory Uncertainties in the apparent gage factor values were found forboth Araldite and 3M DP105-bonded sensor sheets As expected the apparent gage factors for theAraldite adhesive were higher than for the 3M DP105 adhesive We know that an approximate gagefactor for bonding a sensor sheet with 3M DP105 adhesive would allow one to use the sensor sheet toidentify real strain values on structures

The consistency between the strain transfer tests performed on regular strain gages and the testsperformed on sensing sheets confirm that DP105 can be used for the installation of sensing sheets onsteel structural members for crack detection and characterization purposes

Author Contributions BG NV and JES created the concept of sensing sheet MG designed and carried outthe tests and data analysis he is the lead author of the paper CW and LEA developed and manufacturedsensing sheet prototype and associated hardware

Acknowledgments This research was in part supported by the Princeton Institute for the Science andTechnology of Materials (PRISM) and in part by the USDOT-RITA UTC Program grant no DTRT12-G- UTC16enabled through the Center for Advanced Infrastructure and Transportation (CAIT) at Rutgers UniversityThe authors would like to thank Y Hu L Huang N Lin W Rieutort-Louis J Sanz-Robinson T Liu S Wagnerand J Vocaturo from Princeton University for their precious help Kaitlyn Kliever helped improve the contents ofthis paper

Conflicts of Interest The authors declare no conflict of interest

References

1 Yao Y Glišic B Reliable damage detection and localization using direct strain sensing In Proceedings ofthe Sixth International IAMBAS Conference on Bridge Maintenance Safety and Management Stresa Italy8ndash12 July 2012 pp 714ndash721

2 Yao Y Tung S-TE Glišic B Crack detection and characterization techniquesmdashAn overview Struct ControlHealth Monit 2014 21 1387ndash1413 [CrossRef]

3 Tung S-T Yao Y Glišic B Sensing Sheet The Sensitivity of Thin-Film Full- Bridge Strain Sensors for CrackDetection and Characterization Meas Sci Technol 2014 25 14 [CrossRef]

4 Glišic B Chen J Hubbell D Streicker Bridge A comparison between Bragg-gratings long-gaugestrain and temperature sensors and Brillouin scattering-based distributed strain and temperature sensorsIn Proceedings of the SPIE Smart StructuresNDE and Materials + Nondestructive Evaluation and HealthMonitoring San Diego CA USA 6ndash10 March 2011

5 Loh KJ Hou TC Lynch JP Kotov NA Carbon Nanotube Sensing Skins for Spatial Strain and ImpactDamage Identification J Nondestruct Eval 2009 28 9ndash25 [CrossRef]

6 Pyo S Loh KJ Hou TC Jarva E Lynch JP A wireless impedance analyzer for automated tomographicmapping of a nanoengineered sensing skin Smart Struct Syst 2011 8 139ndash155 [CrossRef]

7 Pour-Ghaz M Weiss J Application of Frequency Selective Circuits for Crack Detection in ConcreteElements J ASTM Int 2011 8 1ndash11

8 Schumacher T Thostenson ET Development of structural carbon nanotube-based sensing composites forconcrete structures J Intell Mater Syst Struct 2014 25 1331ndash1339 [CrossRef]

Sensors 2018 18 1907 16 of 16

9 Laflamme S Saleem H Song CH Vasan B Geiger R Chen D Kessler M Bowler N Rajan K SensingSkin for Condition Assessment of Civil Infrastructure In Structural Health Monitoring 2013 A Roadmap toIntelligent Structures Proceedings of the 9th International Workshop on Structural Health Monitoring September10ndash12 2013 DEStech Publications Inc Lancaster PA USA 2013 Volume 2 pp 2795ndash2802

10 Salowitz N Guo Z Kim S-J Li Y-H Lanzara G Chang F-K Screen Printed PiezoceramicActuatorsSensors Microfabricated on Organic Films and Stretchable Networks In Proceedings of the9th International Workshop on Structural Health Monitoring Palo Alto CA USA 10ndash12 September 2013

11 Withey PA Vemuru VSM Bachilo SM Nagarajaiah S Weisman RB Strain paint Noncontact strainmeasurement using single-walled carbon nanotube composite coatings Nano Lett 2012 12 3497ndash3500[CrossRef] [PubMed]

12 Ladani RB Wu S Kinloch AJ Ghorbani K Mouritz AP Wang CH Enhancing fatigue resistance anddamage characterisation in adhesively-bonded composite joints by carbon nanofibres Compos Sci Technol2017 149 116ndash126 [CrossRef]

13 Lim AS Melrose ZR Thostenson ET Chou TW Damage sensing of adhesively-bonded hybridcompositesteel joints using carbon nanotubes Compos Sci Technol 2011 71 1183ndash1189 [CrossRef]

14 Wu S Ladani RB Ravindran AR Zhang J Mouritz AP Kinloch AJ Wang CH Aligning carbonnanofibres in glass-fibreepoxy composites to improve interlaminar toughness and crack-detection capabilityCompos Sci Technol 2017 152 46ndash56 [CrossRef]

15 Zonta D Chiappini A Chiasera A Ferrari M Pozzi M Battisti L Benedetti M Photonic crystals formonitoring fatigue phenomena in steel structures In Proceedings of the SPIE Smart Structures and Materials+ Nondestructive Evaluation and Health Monitoring San Diego CA USA 8ndash12 March 2009 Volume 7292

16 Yao Y Glišic B Detection of steel fatigue cracks with strain sensing sheets based on large area electronicsSensors 2015 15 8088ndash8108 [CrossRef] [PubMed]

17 Glišic B Verma N Very dense arrays of sensors for SHM based on large area electronics In Structural HealthMonitoring 2011 Condition-Based Maintenance and Intelligent Structures Proceedings of the 8th InternationalWorkshop on Structural Health Monitoring Stanford University Stanford CA September 13ndash15 2011 DEStechPublications Inc Lancaster PA USA 2011 Volume 2 pp 1409ndash1416

18 Hu Y Rieutort-Louis WSA Sanz-Robinson J Huang L Glišic B Sturm JC Wagner S Verma NLarge-Scale Sensing System Combining Large-Area Electronics and CMOS ICs for Structural-HealthMonitoring IEEE J Solid-State Circuits 2014 49 1ndash12 [CrossRef]

19 Verma N Hu Y Huang L Rieutort-Louis W Sanz-Robinson J Moy T Glišic B Wagner S SturmJC Enabling Scalable Hybrid Systems Architectures for exploiting large-area electronics in applicationsProc IEEE 2015 103 690ndash712 [CrossRef]

20 Glišic B Yao Y Tung S-T Wagner S Sturm JC Verma N Strain Sensing Sheets for Structural HealthMonitoring based on Large-area Electronics and Integrated Circuits Proc IEEE 2016 104 1513ndash1528[CrossRef]

21 Arias AC MacKenzie JD McCulloch I Rivnay J Salleo A Materials and applications for large areaelectronics Solution-based approaches Chem Rev 2010 110 3ndash24 [CrossRef] [PubMed]

22 Someya T Pal B Huang J Katz HE Organic semiconductor devices with enhanced field andenvironmental responses for novel applications MRS Bull 2008 33 690ndash696 [CrossRef]

23 Someya T Sakurai T Integration of organic field-effect transistors and rubbery pressure sensors for artificialskin applications In Proceedings of the International Electron Devices Meeting Washington DC USA8ndash10 December 2003 pp 203ndash206

24 Subramanian V Lee JB Liu VH Molesam S Printed electronic nose vapor sensors for consumer productmonitoring In Proceedings of the International Solid-State Circuits Conference San Francisco CA USA6ndash9 February 2006 pp 274ndash275

25 Wyrsch N Dunand S Miazza C Shah A Anelli G Despeisse M Garrigos A Jarron P Kaplon JMoraes D et al Thin-film silicon detectors for particle detection Phys Status Solidi 2004 1 1284ndash1291[CrossRef]

ccopy 2018 by the authors Licensee MDPI Basel Switzerland This article is an open accessarticle distributed under the terms and conditions of the Creative Commons Attribution(CC BY) license (httpcreativecommonsorglicensesby40)